Setup details

Author <- "Your Name"
Rev.Title <- "Review Title"
filename.ES.Data<- "data_ostracism.csv"
filename.PRISMA.Data<- "prisma.csv"
X<- "Name of Outcome"
Y <- "Name of Intervention (e.g. CBT)"
Z <- "Name of Control (e.g. Waitlist)"

Data analysis

This file documents the analyses conducted for Your Name (2017) Review Title. Analyses were conducted using the file data_ostracism.csv.

A random-effects meta-analysis (k = 52) was conducted using the REML estimator. The effect size is the standarised mean difference (hedges g).

Results

This review was reported according to the PRISMA statement. The flow of studies through the review is shown in the PRISMA Flowchart below (Figure 1).

Figure 1. PRISMA Flowchart

Results

Figure 2. Forest Plot

Figure 2. Forest Plot

A random effects meta-analysis was conducted (k=52) to explore the difference in Name of Outcome between the Name of Intervention (e.g. CBT) group and the Name of Control (e.g. Waitlist) group. The average difference in Name of Outcome was g=0.13 (p=0.344, 95% CI [-0.14, 0.41]).1 The results of this analysis are summarised in the tables below.

A Cochran’s Q test was conducted to examine whether variations in the observed effect are likely to be attributable soley to sampling error (Q(df=51)=1004.4, p=<.001). The variation in the effect is greater than would be expected from sampling error alone. It appears that the true effect varies betweeen studies. The I2 statistics indicates the proportion of variance in the observed effect attributable to sampling error. In this instance, the I2 = 95.99%.2

Effect Size
g se z p 95% CI LB 95% CI UB
0.133 0.141 0.947 0.344 -0.143 0.41
Heterogeneity
k \(\tau\)2 se Q p I2
52 0.979 0.204 1004.397 0 95.992

A funnel plot was generated to allow for visual inspection of funnel plot asymmetry that can indicate reporting biases (e.g. publication bias or outcome selection bias). An Egger’s test was conducted to detect funnel plot asymmetry (z=3.12,p=0).

Figure 3. Funnel Plot

Figure 3. Funnel Plot

  1. It is important to note that a p>.05 indicates lack of evidence of an effect (i.e. uncertainty) rather than evidence of no effect unless confidence intervals are sufficently narrow to rule out a clinically meaningful effect.↩

  2. Note, this statistic is not an absolute measure of heterogeneity (although it is often interpreted as such). We strongly advise against using rules of thumb such as “small”, “medium” or “large” when interpreting I2 values.↩