Standardized mean differences and study characteristics from the corrected
dataset accompanying Chan and Albarracin's meta-analysis of science-relevant
misinformation. The name Chan25 refers to the 2025 Author Correction.
A data frame with 207 observations on 30 variables and 75 distinct
source PaperNumber identifiers.
ridSource row identifier, stored as character data.
PaperNumberSource report/study grouping identifier, stored as character data; decimal suffixes are retained.
AuthorsAuthors of the source report.
PubYearPublication year of the source report.
ExpNumExperiment number as supplied by the source authors.
SourceTypePublication category: published article (1), working paper (5), dissertation or thesis (6), or unpublished data (7).
RowNoteCondSource notes identifying the experimental condition or comparison represented by the row.
BetweenWithinMethod of effect-size calculation: between-subjects (1) or within-subjects (2).
InPersonCorrectCorrection delivered in person (1) or not (-1).
LabOnlineLaboratory (1) or online (2) setting.
NrowSample size assigned to the source row.
MeanAgeMean participant age in years.
PctFemaleFemale composition as supplied, usually a percentage;
row r_86 is a proportion. See Details.
PctMaleMale composition as supplied, usually a percentage;
row r_86 is a proportion. See Details.
NegMisinfoNeutral (1) or negative (2) misinformation topic.
DetailedCorrectSuccinct (1) or detailed (2) correction.
LikelyFamiliarTopic likely familiar (1) or unfamiliar (-1).
DomainMisinfoMisinformation domain: politics (1), health (2), environment (3), or other (-1).
FictIssueFictitious (1) or real (-1) issue.
IssuePolarizePolitically polarizing (1) or non-polarizing (-1) topic in the country of the experiment.
USOtherUnited States (1) or other-country (2) sample.
AttCongenialCorrection congenial to prior attitudes (1), or incongruent/non-partisan (-1).
es1.gBias-adjusted standardized mean difference for correction, as supplied by the source authors.
es2.gReversed misinformation-persistence effect, as stored in the source data and labelled in the source analysis code.
es1.gvSupplied sampling variance of es1.g.
es2.gvSupplied sampling variance of es2.g.
mis.gBias-adjusted standardized mean difference for misinformation, before imputation.
mis.gvSupplied sampling variance of mis.g.
impmis.gMisinformation effect used in the source analysis including imputed values.
impmis.gvSupplied variance associated with impmis.g;
missing values are retained.
The data retain all rows, variable names, missing values, and numeric values
from the first worksheet of meta-scirelmisinfo_v20240320.xlsx.
No effect sizes, variances, or moderator codes are recalculated or imputed.
Row r_86 stores PctFemale and PctMale as proportions
(0.6081632653 and 0.3918367347), whereas the other rows use percentages.
These entries are retained unchanged. Yousuf et al. (2021), Table 1,
report 596 female participants among 980 in total, confirming that the female
entry is already a proportion. Analyses requiring proportions must account
for this difference in units rather than divide every entry by 100.
The primary report is available at tools:::Rd_expr_doi("10.1016/j.eclinm.2021.100881").
The 207 wide-format rows contain 156 correction effects and 89 reversed misinformation-persistence effects, giving 245 effects in total.
Several rows share a PaperNumber; rows must not be assumed independent.
The identifier is retained as character data. Sampling covariances between
mis.g and es1.g are not supplied. A working correlation used
for robust variance estimation in the source analysis does not establish
this sampling covariance. Sample sizes may recur across rows and should not
be summed as counts of distinct participants.
data(Chan25)
dim(Chan25)
head(Chan25[c("rid", "PaperNumber", "es1.g", "es1.gv", "mis.g", "mis.gv")])
## Complete rows for regression of correction on misinformation
vars <- c("PaperNumber", "es1.g", "es1.gv", "mis.g", "mis.gv")
dat <- Chan25[complete.cases(Chan25[vars]), vars]
nrow(dat)
length(unique(dat$PaperNumber))
Run the code above in your browser using DataLab