SlideShare a Scribd company logo
1 of 1
Download to read offline
On the News Stand: A Meta-Analysis of the Effect of Pretrial Publicity on Guilt
                                                                                                                                                                                                                                                       Sara Marie House
                                                                                                                                                                                                                                                    Loyola University Chicago
                                                Introduction                                                                                                        Methods
                                                                                                                                                                                                                                                                          Results of Pretrial Effect Sizes                                                                                            Exploratory Analyses of Posttrial Effect Sizes
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                     Remedies. Finally, exploratory analyses were performed on remedies. Though none of the remedy
                                                                                                                 Inclusion/Exclusion Criteria
Pretrial publicity (PTP) refers to any news information appearing before a case has gone to trial. It is a                                                                                                                     37 independent effect sizes, with a total N of 7,629 were used. Fixed-effects analysis of the effect sizes yielded an average r̅                                                                                                                                                      variables showed significant effects in the meta-regression, this analysis did not allow the testing of what
cause for concern because it can be biasing to potential jurors. The courts use a variety of remedies to         Any study providing data on a statistical relationship between pretrial publicity (displayed through a mass   = 0.301, SE = 0.011, Z = 28.47, p < 0.0001. The analyses revealed that there was, as expected, substantial variation in effect                                                                                       Strength of evidence. Though the 95% confidence                  Bruschke and Loges (2004) call the cumulative remedy hypothesis, which states that, while a single
                                                                                                                                                                                                                                                                                                                                                                    Strength of Evidence k            N         r          SE        95% CI
counteract the possible effects of pretrial publicity:                                                           media source) and at least one measure of guilt, which could be either dichotomous (not guilty/guilty) or     sizes, Q (df 36) = 457.804, p < 0.0001, I2 (percentage of variance not attributable to sampling error) = 92.14%. Therefore,                                                                                          intervals of weak and strong do overlap, the 90%                 remedy may show no significant effect, the effect of PTP may be reduced through use of a combination of
                                                                                                                 continuous, was considered for meta-analysis. Adequate information about sample characteristics and how       random effects analyses were used:                                                                                                  Weak                    26 2977 0.086* 0.031 0.011, 0.159                        confidence intervals do not, meaning that these two values
•     Continuance – wait for a period of time before beginning the trial                                                                                                                                                                                                                                                                                                                                                                                                                                             remedies. Average effect sizes were computed based on the number of remedies used (ranging from 0 to
•     Extended voir dire – ask more questions of potential jurors                                                the sample was obtained needed to be provided.                                                                                                                                                                                                                                                                                     can be considered significantly different at the 0.10 level.     4) as well as by combination of remedies used.
                                                                                                                                                                                                                                                                                            Analyses revealed no significant moderators (see Meta-                 Moderate                37 6162 0.213* 0.031 0.154, 0.270
•     Admonition – tell jurors to disregard anything learned before trial                                                                                                                                                                                                 Q    df     p                                                                                                                                                             As expected, studies using moderate case evidence had
                                                                                                                 Studies examining the effect of positive PTP were not included in the present meta-analysis.                                                                               Regression Results table). The random-effects component,
                                                                                                                                                                                                                                                                                                                                                                   Strong                  4      1406 0.159              0.089 -0.017, 0.325       the largest effect.
•     Change of venire – use potential jurors from another jurisdiction                                                                                                                                                        Effect of moderators                     3.89    9 0.92      however, explained a significant proportion of variance,. The                                                                                                                                                             Remedies                                                           k     N       r      SE       95% CI
                                                                                                                 Studies could take place in the United States, Canada, or Great Britain. Due to these criteria, all studies
•     Change of venue – move the trial to another jurisdiction                                                                                                                                                                                                                              grand mean effect size was r = 0.323, SE = 0.042, Z = 7.96, p <                                                                                                                                                              None      Overall                                               12 1962 0.159* 0.055 0.054, 0.260
                                                                                                                 obtained were in English.                                                                                     Variance after moderators                22.63 27 0.70
Unfortunately, there is not a lot of evidence in research that these remedies actually work. Many people                                                                                                                                                                                    0.001, 95% CI = 0.245, 0.397, 95% PI = -0.223, 0.893,                 Crime. Crime also appears to Crime                                                               k      N       r     SE        95% CI              1 remedy
associated with the justice system, however, have lamented that current research findings do not provide         Moderators                                                                                                    Overall variance after random-effects 26.52 36 0.88
                                                                                                                                                                                                                                                                                            fail-safe N = 114.                                                                                                                                                                                                                     Admonition                                            10 1101 0.152* 0.063 0.034, 0.266
                                                                                                                                                                                                                                                                                                                                                                  have a strong influence on effect Civil                                                          3      322 0.238* 0.118 0.013, 0.440
information on how pretrial publicity can influence jury verdicts (Chesterman, 1997; Jones, 1991).               •     Source of sample                                                                                                                                                                                                                                                                                                                                                                                            Delay                                                 7    964    0.106 0.077 -0.040, 0.248
                                                                                                                                                                                                                                                                                                                                                                  sizes. Specifically, non-violent        Homicide                                                13 1900 0.290* 0.055 0.193, 0.381
Furthermore, even among pretrial publicity researchers, this issue is a matter of debate. The fact that there                                                                                                                                                                                                                                                                                                                                                                                                                      Deliberation                                          3    464 0.273* 0.105 0.073, 0.452
                                                                                                                 •     Method of assignment to groups                                                                                                                     Results of Posttrial Effect Sizes                                                       crimes and crimes involving             Other Non-Violent                                        7 1047 0.083 0.071 -0.052, 0.215
is so much variation in not only findings but methodology only fuels this debate further. The current study                                                                                                                                                                                                                                                                                                                                                                                                                        Instructions                                          6 1187 0.157* 0.077 0.011, 0.296
                                                                                                                 •     Information contained in the PTP                                                                        67 independent effect sizes, with a total N of 10,545 were used. Fixed-effects analysis of the effect sizes yielded an average r   sexual assault produce smaller          Other Violent                                           16 2943 0.188* 0.045 0.098, 0.275
is one attempt to discover how and when pretrial publicity can bias verdicts, and to answer questions on                                                                                                                                                                                                                                                                                                                                                                                                                           Overall                                               26 3716 0.142* 0.032 0.070, 0.214
                                                                                                                 •     Type of trial – Criminal, Civil                                                                         = 0.163, SE = 0.009, Z = 17.47, p < 0.0001. Once again, the analyses revealed that there was substantial variation in effect       effect sizes. There are two
                                                                                                                                                                                                                                                                                                                                                                                                          Sex Offense                                              9 1349 0.078 0.063 -0.048, 0.201                   2 remedies
whether PTP has an effect posttrial, when it has an effect, and whether remedies work.                                                                                                                                                                                                                                                                            possible explanations for the
                                                                                                                 •     Crime(s) – Homicide, sexual assault, other violent, other nonviolent                                    sizes, Q (df 66) = 378.423, p < 0.0001, I2 = 82.29%.                                                            Q    df      p                                                                                                                                                                      Admonition + Delay                                    8 1160 0.169* 0.063 0.043, 0.291
                                                                                                                                                                                                                                                                                                                                                                                                          Combination of Homicide and Other Crimes 16 2948 0.105* 0.045 0.015, 0.195
Past Meta-Analysis                                                                                               •     Type of media                                                                                           Analyses revealed significant moderators (see Meta-Regression                                                                      smaller effect sizs in sexual                                                                                                                                    Admonition + Deliberation                             1    80     0.139 0.197 -0.244, 0.484
                                                                                                                                                                                                                                                                                                    Effect of moderators                    25.16 17 0.006                                                Unknown Combination                                      4      202   0.159 0.110 -0.053, 0.357
                                                                                                                                                                                                                               Results table). The grand mean effect size was r = 0.163,                                                                          assault cases: 1) PTP publishes                                                                                                                                  Admonition + Instructions                             1    202    0.323 0.176 -0.012, 0.593
The most recent, and only published, meta-analysis was performed by Steblay, Besirevic, Fulero, and              •     Stimuli provided to control group
                                                                                                                                                                                                                                                                                                    Variance after moderators               52.36 49 0.35         negative information about the victim, and 2) jurors are unsympathetic to victims of sexual assault. Since sexual                                                Delay + Deliberation
                                                                                                                                                                                                                               SE = 0.023, Z = 7.14, p < 0.001, 95% CI = 0.118, 0.207,                                                                                                                                                                                                                                                                                                   11 1911 0.166* 0.055 0.054, 0.274
Jimenz-Lorente (1999). They used 44 independent effect sizes with an overall sample size of 5,755.               •     Remedies
                                                                                                                                                                                                                               95% PI = -0.222, 0.550, fail-safe N = 178. The posttrial             Overall variance after random-effects 87.52 66 0.04           assault was not a significant moderator pretrial, but was posttrial, the evidence points to explanation 2.                                                       Delay + Instructions                                  1    168   -0.035 0.179 -0.369, 0.308
Using a fixed-effects model, Steblay, et al., found a mean effect size of 0.16 (95% CI = -0.13, 0.46)            •    Strength of evidence – inferred in one of three ways: 1) statement by author of case strength, 2)
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   Deliberation + Voi r Dire                             1    156    0.111 0.182 -0.239, 0.436
                                                                                                                 conviction rate from pilot group, or 3) conviction rate from control group; less than 35% = weak, 35 to       mean effect size was smaller than the pretrial mean effect size. In addition, the 95% confidence intervals do not overlap,
Several moderators were examined: study design, source of sample, time of verdict, delay between                                                                                                                                                                                                                                                                                                                                                                                                                                   Overall                                               23 3677 0.175* 0.045 0.099, 0.248
                                                                                                                 69% = moderate, and 70% or higher = strong                                                                    meaning that these two values can be considered significantly different at the 0.05 level. This confirms the ‘timing                                                                                                     Trial presentation method. Finally, since trial
exposure to PTP and verdict, content of PTP, specificity (case-specific versus general), crime, and media                                                                                                                                                                                                                                                         Trial Presentation Method       k       N           r         SE     95% CI                                                                         3 remedies
                                                                                                                 •     Characteristics of participants (gender, race, etc.) – dropped due to missingness                       hypothesis’.                                                                                                                                                                                                             presentation method was found to be a significant
(newspaper, video, or both). Each moderator was analyzed individually; all effect sizes were categorized                                                                                                                                                                                                                                                          Brief Summary                   15 2234 0.147* 0.045 0.053, 0.239                                                                                                Admonition + Delay + Deliberation                     3    372    0.223 0.122 -0.013, 0.435
                                                                                                                 •     Characteristics of defendant – dropped due to missingness                                                                                                                                                                                                                                                                        moderator in the regression, average effect sizes were                     Delay + Deliberation + Voi r Dire                     1    68     0.228 0.205 -0.167, 0.559
on the moderator of interest, and average effect sizes were calculated at each level. Z-tests were performed                                                                                                                                                                                                                                                      Transcript                      11 1603 0.171* 0.055                0.058, .279
                                                                                                                 •     Characteristics of victim – dropped due to missingness                                                       Meta-Regression Results                                                                                                                                                                                             computed at each level of presentation. Studies using                      Overall                                               4    440 0.248* 0.100 0.058, 0.420
on these averages to test whether they were significantly larger than 0. Unfortunately, this method did not                                                                                                                                                                                                  Pretrial Verdict         Posttrial Verdict
                                                                                                                                                                                                                                                                                                                                                                  Audio                           7 1145 0.154* 0.071 0.020, 0.283                      mock trials or actual trials did not find effect sizes
                                                                                                                                                                                                                                    Variable                                                                                                                                                                                                                                                                          4 remedies
allow significance testing on differences between averages, nor did it allow all moderators to be analyzed       Coding                                                                                                                                                                                  B           95% CI         B          95% CI
                                                                                                                                                                                                                                                                                                                                                                  Video                           25 4234 0.194* 0.032 0.123, 0.263                     significantly larger than 0, though these effect sizes are                 Admonition + Delay + Deliberation + Instructions      1    702    0.246 0.167 -0.076, 0.521
at once.                                                                                                         Coding of reports was performed by the study author. Interrater reliability is currently being assessed; no        Constant                                                          0.642      -0.174, 1.457   0.013     -0.457, 0.484
                                                                                                                                                                                                                                                                                                                                                                  Mock Trial                      3       144       0.037 0.138 -0.227, 0.295           also based on only a few studies, and could suffer from                    Admonition + Delay + Deliberation + Voir Dire         1    48       0     0.221 -0.407, 0.407
The current meta-analysis uses 104 effect sizes from 99 independent samples; using weighted least squares        information is available at this time.                                                                             Sample                                                            -0.064 -0.194, 0.065       0.006     -0.060, 0.072                                                                                                                                                                           Overall                                               2    750    0.160 0.134 -0.100, 0.397
                                                                                                                                                                                                                                                                                                                                                                  Actual Trial                    6 1185 0.095 0.084 -0.065, 0.250                      low power.
multiple regression, multiple moderators were analyzed at once to determine a moderator’s ability to                                                                                                                                Media Type                                                        -0.024 -0.370, 0.322       -0.033    -0.176, 0.111
                                                                                                                 Statistical Methods
account for unique variance. In addition, the present meta-analysis used random-effects models for all                                                                                                                              PTP Information                                                   -0.067 -0.309, 0.175       0.142     -0.032, 0.316
                                                                                                                 Effects sizes were computed as Pearson’s r, which were converted using Fisher’s z-transformation prior to
analyses, which Borenstein, et al. (2009) argue tends to be more theoretically justified.
                                                                                                                 analysis. Results were then converted back to Pearson’s r.                                                         Control Group Instructions                                        -0.045 -0.184, 0.094       0.033     -0.013, 0.079                                                                                                                                             References
Hypotheses
                                                                                                                                                                                                                                    Homicide                                                          0.041      -0.169, 0.250   0.117     -0.013, 0.247
Timing Hypothesis: Pretrial effect sizes will be larger than posttrial effect sizes.                             If a study had more than one group or more than one measure of the dependent variable, information was                                                                                                                                           Sources used in the meta-analysis are available at: http://saramhouse.bravehost.com/research/ptpmetaanalysis.html
                                                                                                                 combined to create a single effect size. Variances for these averaged effect sizes were recalculated using         Sex Offense                                                       -0.003 -0.227, 0.221       -0.178* -0.325, -0.031           Borenstein, M. Hedges, L.V., Higgins, J.P.T., & Rothstein, H.R. (2009). Introduction to meta-analysis. West Sussex, UK: John Wiley & Sons.
Length Hypothesis: Longer trial materials will produce smaller effect sizes.                                                                                                                                                                                                                                                                                      Bruschke, J., & Loges, W.E. (2004). Free press vs. fair trials: Examining publicity’s role in trial outcomes. Mahwah, NJ: Lawrence Erlbaum.
                                                                                                                 the procedure provided by Borenstein, et al. (2009).                                                               Other Violent Crime                                               -.0137 -0.366, 0.093       -0.056    -0.161, 0.049
Amount Hypothesis: Studies using larger amounts of PTP will have larger effect sizes.                                                                                                                                                                                                                                                                             Chesterman, M. (1997). OJ and the dingo: How media publicity relating to criminal cases tried by jury is dealt with in Australia and America. The American Journal of Comparative Law, 45, 109-147.
                                                                                                                 Time of measurement was expected to have an effect, so these values were not averaged together and                 Other Non-Violent Crime                                           0.023      -0.241, 0.288   -0.103    -0.297, 0.090
                                                                                                                                                                                                                                                                                                                                                                  Cohn, L.D., & Becker, B.J. (2003). How meta-analysis increases statistical power. Psychological Methods, 8, 243-253. doi:10.1037/1082-989X.8.3.243
Strength Hypothesis: Studies using cases with moderate-strength evidence will produce larger effect sizes        pretrial were analyzed separately from those measured posttrial.                                                   Delay                                                             -0.053 -0.392, 0.286       -0.055    -0.197, 0.086          Cooper, H. (2010). Research synthesis and meta-analysis: A step-by-step approach (4th ed.). Thousand Oaks, CA: SAGE Publications.
than studies using weak or strong evidence.
                                                                                                                                                                                                                                    Strength of evidence - moderate compared to weak and strong                                  0.198*    0.069, 0.327           Curtner, R., & Kassier, M. (2005). “Not in our town”: Pretrial publicity, presumed prejudice, and change of venue in Alaska: Public opinion surveys as a tool to measure the impact of prejudicial pretrial publicity. Alaska Law Review, 22, 255-
                                                                                                                 Random-effects models were used in both cases. A random-effects model assumes that there is not one
Sample Hypothesis: Studies using student samples will have similar results as studies using non-student                                                                                                                                                                                                                                                              292.
                                                                                                                 true effect size, but rather a distribution of effect sizes, based on a variety of study characteristics           Strength of evidence - strong compared to weak and moderate                                  0.034     -0.191, 0.259
samples.                                                                                                                                                                                                                                                                                                                                                          Dixon, T.L., & Linz, D. (2002). Television news, prejudicial pretrial publicity, and the depiction of race. Journal of Broadcasting & Electronic Media, 46, 112-136.
                                                                                                                 (Borenstein, et al., 2009; Cooper, 2010; Lipsey & Wilson, 2001). In order to incorporate multiple                  Deliberation                                                                                 0.109     -0.027, 0.246          Hedges, L.V., & Olkin, I. (1985). Statistical methods for meta-analysis. Orlando, FL: Academic Press.
Remedy Hypothesis: Remedies will reduce the effect of PTP on guilt.
                                                                                                                 moderators into the analyses, weighted least squares regression (also referred to as meta-regression:              Trial presentation format                                                                    -0.030* -0.059, -0.0004          Imrich, D.J., Mullin, C., & Linz, D. (1995). Measuring the extent of prejudicial pretrial publicity in major American newspapers: A content analysis. Journal of Communication, 45, 94-117.
Pretesting Hypothesis: Studies using pretesting will have larger posttrial effect sizes than studies not using   Borenstein, et al., 2009) with methods of moments estimation was performed (based on the SPSS macros                                                                                                                                             Jones, R.M. (1991). The latest empirical studies on pretrial publicity, jury bias, and judicial remedies: Not enough to overcome the first amendment right of access to pretrial hearings. American University Law Review, 40, 841-.
pretesting.                                                                                                                                                                                                                         Judicial admonition                                                                          0.057     -0.068, 0.182
                                                                                                                 provided by Wilson, 2002); the procedure allows for a random intercept but uses fixed slopes. This                                                                                                                                               Lipsey, M.W., & Wilson, D.B. (2001). Practical meta-analysis. Thousand Oaks, CA: SAGE Publications.
                                                                                                                                                                                                                                    Judicial instructions                                                                        -0.107    -0.256, 0.043          Steblay, N.M., Besirevic, J., Fulero, S.M., & Jimenz-Lorente, B. (1999). The effects of pretrial publicity on juror verdicts: A meta-analytic review. Law and Human Behavior, 23, 219-235. doi:10.1023/A:1022325019080
Other variables were expected to have some effect, but it was unclear from past research and theory how          analysis enters moderators first, then computes the random error component based on the remaining
these variables might influence results. These analyses are exploratory.                                                                                                                                                            Voir dire                                                                                    -0.061    -0.271, 0.150          Studebaker, C.A., & Penrod, S.D. (2005). Pretrial publicity and its influence on juror decision making. In N. Brewer and K.D. Williams (Eds.), Psychology and law: An empirical perspective (pp. 254-275). New York: Guilford Press.
                                                                                                                 variance (Lipsey & Wilson, 2001).
                                                                                                                                                                                                                                                                                                                                                                  Wilson, D.B. (2002). Meta-analysis macros for SAS, SPSS, and Stata. Retrieved September 9, 2009, from http://mason.gmu.edu/~dwilsonb/ma.html
                                                                                                                                                                                                                                    Pretesting                                                                                   -0.080    -0.201, 0.042

More Related Content

Similar to APS 2010 Poster

Sample-size-comprehensive.pptx
Sample-size-comprehensive.pptxSample-size-comprehensive.pptx
Sample-size-comprehensive.pptxssuser4eb7dd
 
How Randomized Controlled Trials are Used in Meta-Analysis
How Randomized Controlled Trials are Used in Meta-Analysis How Randomized Controlled Trials are Used in Meta-Analysis
How Randomized Controlled Trials are Used in Meta-Analysis Pubrica
 
Minimizing Risk In Phase II and III Sample Size Calculation
Minimizing Risk In Phase II and III Sample Size CalculationMinimizing Risk In Phase II and III Sample Size Calculation
Minimizing Risk In Phase II and III Sample Size CalculationnQuery
 
Research and Scientific Journal Publication support services | Research pape...
 Research and Scientific Journal Publication support services | Research pape... Research and Scientific Journal Publication support services | Research pape...
Research and Scientific Journal Publication support services | Research pape...Pubrica
 
Sample determinants and size
Sample determinants and sizeSample determinants and size
Sample determinants and sizeTarek Tawfik Amin
 
Sample size
Sample sizeSample size
Sample sizezubis
 
Bayesian random effects meta-analysis model for normal data - Pubrica
Bayesian random effects meta-analysis model for normal data - PubricaBayesian random effects meta-analysis model for normal data - Pubrica
Bayesian random effects meta-analysis model for normal data - PubricaPubrica
 
Response of Watermelon to Five Different Rates of Poultry Manure in Asaba Are...
Response of Watermelon to Five Different Rates of Poultry Manure in Asaba Are...Response of Watermelon to Five Different Rates of Poultry Manure in Asaba Are...
Response of Watermelon to Five Different Rates of Poultry Manure in Asaba Are...IOSR Journals
 
STATISTICS : Changing the way we do: Hypothesis testing, effect size, power, ...
STATISTICS : Changing the way we do: Hypothesis testing, effect size, power, ...STATISTICS : Changing the way we do: Hypothesis testing, effect size, power, ...
STATISTICS : Changing the way we do: Hypothesis testing, effect size, power, ...Musfera Nara Vadia
 
Infosheet east-bayesian-power-calculations
Infosheet east-bayesian-power-calculationsInfosheet east-bayesian-power-calculations
Infosheet east-bayesian-power-calculationsCytel USA
 
East Bayesian Assurance Infosheet
East Bayesian Assurance InfosheetEast Bayesian Assurance Infosheet
East Bayesian Assurance InfosheetCytel
 
East bayesian power calculations
East bayesian power calculationsEast bayesian power calculations
East bayesian power calculationsCytel
 
Meta-Analysis -- Introduction.pptx
Meta-Analysis -- Introduction.pptxMeta-Analysis -- Introduction.pptx
Meta-Analysis -- Introduction.pptxACSRM
 
Practical Methods To Overcome Sample Size Challenges
Practical Methods To Overcome Sample Size ChallengesPractical Methods To Overcome Sample Size Challenges
Practical Methods To Overcome Sample Size ChallengesnQuery
 
Avoid overfitting in precision medicine: How to use cross-validation to relia...
Avoid overfitting in precision medicine: How to use cross-validation to relia...Avoid overfitting in precision medicine: How to use cross-validation to relia...
Avoid overfitting in precision medicine: How to use cross-validation to relia...Nicole Krämer
 
Statistics in meta analysis
Statistics in meta analysisStatistics in meta analysis
Statistics in meta analysisDr Shri Sangle
 
Bayesian Assurance: Formalizing Sensitivity Analysis For Sample Size
Bayesian Assurance: Formalizing Sensitivity Analysis For Sample SizeBayesian Assurance: Formalizing Sensitivity Analysis For Sample Size
Bayesian Assurance: Formalizing Sensitivity Analysis For Sample SizenQuery
 

Similar to APS 2010 Poster (20)

Sample-size-comprehensive.pptx
Sample-size-comprehensive.pptxSample-size-comprehensive.pptx
Sample-size-comprehensive.pptx
 
How Randomized Controlled Trials are Used in Meta-Analysis
How Randomized Controlled Trials are Used in Meta-Analysis How Randomized Controlled Trials are Used in Meta-Analysis
How Randomized Controlled Trials are Used in Meta-Analysis
 
Minimizing Risk In Phase II and III Sample Size Calculation
Minimizing Risk In Phase II and III Sample Size CalculationMinimizing Risk In Phase II and III Sample Size Calculation
Minimizing Risk In Phase II and III Sample Size Calculation
 
Research and Scientific Journal Publication support services | Research pape...
 Research and Scientific Journal Publication support services | Research pape... Research and Scientific Journal Publication support services | Research pape...
Research and Scientific Journal Publication support services | Research pape...
 
Sample determinants and size
Sample determinants and sizeSample determinants and size
Sample determinants and size
 
Sample size
Sample sizeSample size
Sample size
 
Bayesian random effects meta-analysis model for normal data - Pubrica
Bayesian random effects meta-analysis model for normal data - PubricaBayesian random effects meta-analysis model for normal data - Pubrica
Bayesian random effects meta-analysis model for normal data - Pubrica
 
Response of Watermelon to Five Different Rates of Poultry Manure in Asaba Are...
Response of Watermelon to Five Different Rates of Poultry Manure in Asaba Are...Response of Watermelon to Five Different Rates of Poultry Manure in Asaba Are...
Response of Watermelon to Five Different Rates of Poultry Manure in Asaba Are...
 
STATISTICS : Changing the way we do: Hypothesis testing, effect size, power, ...
STATISTICS : Changing the way we do: Hypothesis testing, effect size, power, ...STATISTICS : Changing the way we do: Hypothesis testing, effect size, power, ...
STATISTICS : Changing the way we do: Hypothesis testing, effect size, power, ...
 
Infosheet east-bayesian-power-calculations
Infosheet east-bayesian-power-calculationsInfosheet east-bayesian-power-calculations
Infosheet east-bayesian-power-calculations
 
East Bayesian Assurance Infosheet
East Bayesian Assurance InfosheetEast Bayesian Assurance Infosheet
East Bayesian Assurance Infosheet
 
East bayesian power calculations
East bayesian power calculationsEast bayesian power calculations
East bayesian power calculations
 
Meta-Analysis -- Introduction.pptx
Meta-Analysis -- Introduction.pptxMeta-Analysis -- Introduction.pptx
Meta-Analysis -- Introduction.pptx
 
Practical Methods To Overcome Sample Size Challenges
Practical Methods To Overcome Sample Size ChallengesPractical Methods To Overcome Sample Size Challenges
Practical Methods To Overcome Sample Size Challenges
 
Avoid overfitting in precision medicine: How to use cross-validation to relia...
Avoid overfitting in precision medicine: How to use cross-validation to relia...Avoid overfitting in precision medicine: How to use cross-validation to relia...
Avoid overfitting in precision medicine: How to use cross-validation to relia...
 
Statistics in meta analysis
Statistics in meta analysisStatistics in meta analysis
Statistics in meta analysis
 
Methodology
MethodologyMethodology
Methodology
 
Landau dunn 20march2013
Landau dunn 20march2013Landau dunn 20march2013
Landau dunn 20march2013
 
Bayesian Assurance: Formalizing Sensitivity Analysis For Sample Size
Bayesian Assurance: Formalizing Sensitivity Analysis For Sample SizeBayesian Assurance: Formalizing Sensitivity Analysis For Sample Size
Bayesian Assurance: Formalizing Sensitivity Analysis For Sample Size
 
Sample and effect size
Sample and effect sizeSample and effect size
Sample and effect size
 

APS 2010 Poster

  • 1. On the News Stand: A Meta-Analysis of the Effect of Pretrial Publicity on Guilt Sara Marie House Loyola University Chicago Introduction Methods Results of Pretrial Effect Sizes Exploratory Analyses of Posttrial Effect Sizes Remedies. Finally, exploratory analyses were performed on remedies. Though none of the remedy Inclusion/Exclusion Criteria Pretrial publicity (PTP) refers to any news information appearing before a case has gone to trial. It is a 37 independent effect sizes, with a total N of 7,629 were used. Fixed-effects analysis of the effect sizes yielded an average r̅ variables showed significant effects in the meta-regression, this analysis did not allow the testing of what cause for concern because it can be biasing to potential jurors. The courts use a variety of remedies to Any study providing data on a statistical relationship between pretrial publicity (displayed through a mass = 0.301, SE = 0.011, Z = 28.47, p < 0.0001. The analyses revealed that there was, as expected, substantial variation in effect Strength of evidence. Though the 95% confidence Bruschke and Loges (2004) call the cumulative remedy hypothesis, which states that, while a single Strength of Evidence k N r SE 95% CI counteract the possible effects of pretrial publicity: media source) and at least one measure of guilt, which could be either dichotomous (not guilty/guilty) or sizes, Q (df 36) = 457.804, p < 0.0001, I2 (percentage of variance not attributable to sampling error) = 92.14%. Therefore, intervals of weak and strong do overlap, the 90% remedy may show no significant effect, the effect of PTP may be reduced through use of a combination of continuous, was considered for meta-analysis. Adequate information about sample characteristics and how random effects analyses were used: Weak 26 2977 0.086* 0.031 0.011, 0.159 confidence intervals do not, meaning that these two values • Continuance – wait for a period of time before beginning the trial remedies. Average effect sizes were computed based on the number of remedies used (ranging from 0 to • Extended voir dire – ask more questions of potential jurors the sample was obtained needed to be provided. can be considered significantly different at the 0.10 level. 4) as well as by combination of remedies used. Analyses revealed no significant moderators (see Meta- Moderate 37 6162 0.213* 0.031 0.154, 0.270 • Admonition – tell jurors to disregard anything learned before trial Q df p As expected, studies using moderate case evidence had Studies examining the effect of positive PTP were not included in the present meta-analysis. Regression Results table). The random-effects component, Strong 4 1406 0.159 0.089 -0.017, 0.325 the largest effect. • Change of venire – use potential jurors from another jurisdiction Effect of moderators 3.89 9 0.92 however, explained a significant proportion of variance,. The Remedies k N r SE 95% CI Studies could take place in the United States, Canada, or Great Britain. Due to these criteria, all studies • Change of venue – move the trial to another jurisdiction grand mean effect size was r = 0.323, SE = 0.042, Z = 7.96, p < None Overall 12 1962 0.159* 0.055 0.054, 0.260 obtained were in English. Variance after moderators 22.63 27 0.70 Unfortunately, there is not a lot of evidence in research that these remedies actually work. Many people 0.001, 95% CI = 0.245, 0.397, 95% PI = -0.223, 0.893, Crime. Crime also appears to Crime k N r SE 95% CI 1 remedy associated with the justice system, however, have lamented that current research findings do not provide Moderators Overall variance after random-effects 26.52 36 0.88 fail-safe N = 114. Admonition 10 1101 0.152* 0.063 0.034, 0.266 have a strong influence on effect Civil 3 322 0.238* 0.118 0.013, 0.440 information on how pretrial publicity can influence jury verdicts (Chesterman, 1997; Jones, 1991). • Source of sample Delay 7 964 0.106 0.077 -0.040, 0.248 sizes. Specifically, non-violent Homicide 13 1900 0.290* 0.055 0.193, 0.381 Furthermore, even among pretrial publicity researchers, this issue is a matter of debate. The fact that there Deliberation 3 464 0.273* 0.105 0.073, 0.452 • Method of assignment to groups Results of Posttrial Effect Sizes crimes and crimes involving Other Non-Violent 7 1047 0.083 0.071 -0.052, 0.215 is so much variation in not only findings but methodology only fuels this debate further. The current study Instructions 6 1187 0.157* 0.077 0.011, 0.296 • Information contained in the PTP 67 independent effect sizes, with a total N of 10,545 were used. Fixed-effects analysis of the effect sizes yielded an average r sexual assault produce smaller Other Violent 16 2943 0.188* 0.045 0.098, 0.275 is one attempt to discover how and when pretrial publicity can bias verdicts, and to answer questions on Overall 26 3716 0.142* 0.032 0.070, 0.214 • Type of trial – Criminal, Civil = 0.163, SE = 0.009, Z = 17.47, p < 0.0001. Once again, the analyses revealed that there was substantial variation in effect effect sizes. There are two Sex Offense 9 1349 0.078 0.063 -0.048, 0.201 2 remedies whether PTP has an effect posttrial, when it has an effect, and whether remedies work. possible explanations for the • Crime(s) – Homicide, sexual assault, other violent, other nonviolent sizes, Q (df 66) = 378.423, p < 0.0001, I2 = 82.29%. Q df p Admonition + Delay 8 1160 0.169* 0.063 0.043, 0.291 Combination of Homicide and Other Crimes 16 2948 0.105* 0.045 0.015, 0.195 Past Meta-Analysis • Type of media Analyses revealed significant moderators (see Meta-Regression smaller effect sizs in sexual Admonition + Deliberation 1 80 0.139 0.197 -0.244, 0.484 Effect of moderators 25.16 17 0.006 Unknown Combination 4 202 0.159 0.110 -0.053, 0.357 Results table). The grand mean effect size was r = 0.163, assault cases: 1) PTP publishes Admonition + Instructions 1 202 0.323 0.176 -0.012, 0.593 The most recent, and only published, meta-analysis was performed by Steblay, Besirevic, Fulero, and • Stimuli provided to control group Variance after moderators 52.36 49 0.35 negative information about the victim, and 2) jurors are unsympathetic to victims of sexual assault. Since sexual Delay + Deliberation SE = 0.023, Z = 7.14, p < 0.001, 95% CI = 0.118, 0.207, 11 1911 0.166* 0.055 0.054, 0.274 Jimenz-Lorente (1999). They used 44 independent effect sizes with an overall sample size of 5,755. • Remedies 95% PI = -0.222, 0.550, fail-safe N = 178. The posttrial Overall variance after random-effects 87.52 66 0.04 assault was not a significant moderator pretrial, but was posttrial, the evidence points to explanation 2. Delay + Instructions 1 168 -0.035 0.179 -0.369, 0.308 Using a fixed-effects model, Steblay, et al., found a mean effect size of 0.16 (95% CI = -0.13, 0.46) • Strength of evidence – inferred in one of three ways: 1) statement by author of case strength, 2) Deliberation + Voi r Dire 1 156 0.111 0.182 -0.239, 0.436 conviction rate from pilot group, or 3) conviction rate from control group; less than 35% = weak, 35 to mean effect size was smaller than the pretrial mean effect size. In addition, the 95% confidence intervals do not overlap, Several moderators were examined: study design, source of sample, time of verdict, delay between Overall 23 3677 0.175* 0.045 0.099, 0.248 69% = moderate, and 70% or higher = strong meaning that these two values can be considered significantly different at the 0.05 level. This confirms the ‘timing Trial presentation method. Finally, since trial exposure to PTP and verdict, content of PTP, specificity (case-specific versus general), crime, and media Trial Presentation Method k N r SE 95% CI 3 remedies • Characteristics of participants (gender, race, etc.) – dropped due to missingness hypothesis’. presentation method was found to be a significant (newspaper, video, or both). Each moderator was analyzed individually; all effect sizes were categorized Brief Summary 15 2234 0.147* 0.045 0.053, 0.239 Admonition + Delay + Deliberation 3 372 0.223 0.122 -0.013, 0.435 • Characteristics of defendant – dropped due to missingness moderator in the regression, average effect sizes were Delay + Deliberation + Voi r Dire 1 68 0.228 0.205 -0.167, 0.559 on the moderator of interest, and average effect sizes were calculated at each level. Z-tests were performed Transcript 11 1603 0.171* 0.055 0.058, .279 • Characteristics of victim – dropped due to missingness Meta-Regression Results computed at each level of presentation. Studies using Overall 4 440 0.248* 0.100 0.058, 0.420 on these averages to test whether they were significantly larger than 0. Unfortunately, this method did not Pretrial Verdict Posttrial Verdict Audio 7 1145 0.154* 0.071 0.020, 0.283 mock trials or actual trials did not find effect sizes Variable 4 remedies allow significance testing on differences between averages, nor did it allow all moderators to be analyzed Coding B 95% CI B 95% CI Video 25 4234 0.194* 0.032 0.123, 0.263 significantly larger than 0, though these effect sizes are Admonition + Delay + Deliberation + Instructions 1 702 0.246 0.167 -0.076, 0.521 at once. Coding of reports was performed by the study author. Interrater reliability is currently being assessed; no Constant 0.642 -0.174, 1.457 0.013 -0.457, 0.484 Mock Trial 3 144 0.037 0.138 -0.227, 0.295 also based on only a few studies, and could suffer from Admonition + Delay + Deliberation + Voir Dire 1 48 0 0.221 -0.407, 0.407 The current meta-analysis uses 104 effect sizes from 99 independent samples; using weighted least squares information is available at this time. Sample -0.064 -0.194, 0.065 0.006 -0.060, 0.072 Overall 2 750 0.160 0.134 -0.100, 0.397 Actual Trial 6 1185 0.095 0.084 -0.065, 0.250 low power. multiple regression, multiple moderators were analyzed at once to determine a moderator’s ability to Media Type -0.024 -0.370, 0.322 -0.033 -0.176, 0.111 Statistical Methods account for unique variance. In addition, the present meta-analysis used random-effects models for all PTP Information -0.067 -0.309, 0.175 0.142 -0.032, 0.316 Effects sizes were computed as Pearson’s r, which were converted using Fisher’s z-transformation prior to analyses, which Borenstein, et al. (2009) argue tends to be more theoretically justified. analysis. Results were then converted back to Pearson’s r. Control Group Instructions -0.045 -0.184, 0.094 0.033 -0.013, 0.079 References Hypotheses Homicide 0.041 -0.169, 0.250 0.117 -0.013, 0.247 Timing Hypothesis: Pretrial effect sizes will be larger than posttrial effect sizes. If a study had more than one group or more than one measure of the dependent variable, information was Sources used in the meta-analysis are available at: http://saramhouse.bravehost.com/research/ptpmetaanalysis.html combined to create a single effect size. Variances for these averaged effect sizes were recalculated using Sex Offense -0.003 -0.227, 0.221 -0.178* -0.325, -0.031 Borenstein, M. Hedges, L.V., Higgins, J.P.T., & Rothstein, H.R. (2009). Introduction to meta-analysis. West Sussex, UK: John Wiley & Sons. Length Hypothesis: Longer trial materials will produce smaller effect sizes. Bruschke, J., & Loges, W.E. (2004). Free press vs. fair trials: Examining publicity’s role in trial outcomes. Mahwah, NJ: Lawrence Erlbaum. the procedure provided by Borenstein, et al. (2009). Other Violent Crime -.0137 -0.366, 0.093 -0.056 -0.161, 0.049 Amount Hypothesis: Studies using larger amounts of PTP will have larger effect sizes. Chesterman, M. (1997). OJ and the dingo: How media publicity relating to criminal cases tried by jury is dealt with in Australia and America. The American Journal of Comparative Law, 45, 109-147. Time of measurement was expected to have an effect, so these values were not averaged together and Other Non-Violent Crime 0.023 -0.241, 0.288 -0.103 -0.297, 0.090 Cohn, L.D., & Becker, B.J. (2003). How meta-analysis increases statistical power. Psychological Methods, 8, 243-253. doi:10.1037/1082-989X.8.3.243 Strength Hypothesis: Studies using cases with moderate-strength evidence will produce larger effect sizes pretrial were analyzed separately from those measured posttrial. Delay -0.053 -0.392, 0.286 -0.055 -0.197, 0.086 Cooper, H. (2010). Research synthesis and meta-analysis: A step-by-step approach (4th ed.). Thousand Oaks, CA: SAGE Publications. than studies using weak or strong evidence. Strength of evidence - moderate compared to weak and strong 0.198* 0.069, 0.327 Curtner, R., & Kassier, M. (2005). “Not in our town”: Pretrial publicity, presumed prejudice, and change of venue in Alaska: Public opinion surveys as a tool to measure the impact of prejudicial pretrial publicity. Alaska Law Review, 22, 255- Random-effects models were used in both cases. A random-effects model assumes that there is not one Sample Hypothesis: Studies using student samples will have similar results as studies using non-student 292. true effect size, but rather a distribution of effect sizes, based on a variety of study characteristics Strength of evidence - strong compared to weak and moderate 0.034 -0.191, 0.259 samples. Dixon, T.L., & Linz, D. (2002). Television news, prejudicial pretrial publicity, and the depiction of race. Journal of Broadcasting & Electronic Media, 46, 112-136. (Borenstein, et al., 2009; Cooper, 2010; Lipsey & Wilson, 2001). In order to incorporate multiple Deliberation 0.109 -0.027, 0.246 Hedges, L.V., & Olkin, I. (1985). Statistical methods for meta-analysis. Orlando, FL: Academic Press. Remedy Hypothesis: Remedies will reduce the effect of PTP on guilt. moderators into the analyses, weighted least squares regression (also referred to as meta-regression: Trial presentation format -0.030* -0.059, -0.0004 Imrich, D.J., Mullin, C., & Linz, D. (1995). Measuring the extent of prejudicial pretrial publicity in major American newspapers: A content analysis. Journal of Communication, 45, 94-117. Pretesting Hypothesis: Studies using pretesting will have larger posttrial effect sizes than studies not using Borenstein, et al., 2009) with methods of moments estimation was performed (based on the SPSS macros Jones, R.M. (1991). The latest empirical studies on pretrial publicity, jury bias, and judicial remedies: Not enough to overcome the first amendment right of access to pretrial hearings. American University Law Review, 40, 841-. pretesting. Judicial admonition 0.057 -0.068, 0.182 provided by Wilson, 2002); the procedure allows for a random intercept but uses fixed slopes. This Lipsey, M.W., & Wilson, D.B. (2001). Practical meta-analysis. Thousand Oaks, CA: SAGE Publications. Judicial instructions -0.107 -0.256, 0.043 Steblay, N.M., Besirevic, J., Fulero, S.M., & Jimenz-Lorente, B. (1999). The effects of pretrial publicity on juror verdicts: A meta-analytic review. Law and Human Behavior, 23, 219-235. doi:10.1023/A:1022325019080 Other variables were expected to have some effect, but it was unclear from past research and theory how analysis enters moderators first, then computes the random error component based on the remaining these variables might influence results. These analyses are exploratory. Voir dire -0.061 -0.271, 0.150 Studebaker, C.A., & Penrod, S.D. (2005). Pretrial publicity and its influence on juror decision making. In N. Brewer and K.D. Williams (Eds.), Psychology and law: An empirical perspective (pp. 254-275). New York: Guilford Press. variance (Lipsey & Wilson, 2001). Wilson, D.B. (2002). Meta-analysis macros for SAS, SPSS, and Stata. Retrieved September 9, 2009, from http://mason.gmu.edu/~dwilsonb/ma.html Pretesting -0.080 -0.201, 0.042