Desktop version

Home arrow Economics

  • Increase font
  • Decrease font


<<   CONTENTS   >>

Analyzing Ten Hypothetical Trials

Suppose that 10 estimates for the treatment effects on S and T are available in the published literature, as well as the sample sizes on which these estimates were based. Using this information, an estimate of Rrial can be obtained using the following command:

# Fit the model

> Trial_Fit <- TrialLevelMA(

Alpha.Vector=c(4.7, 4.9, 5.2, 5.7, 5.1, 5.8, 6.0, 5.8, 5.9,

  • 5.4), Beta.Vector=c(13.6, 15.3, 15.9, 16.4, 16.1, 18.5, 17.3, 18.2, 17.7, 16.4), N.Vector=c(130, 140, 150, 200, 210, 240,
  • 300, 350, 350, 400))

The fitted object Trial_Fit of class TrialLevelMA can subsequently be examined by applying the summary() and plot() functions:

# Obtain summary of the results:

> summary(Trial_Fit)

Function call:

TrialLevelMA(Alpha.Vector = c(4.7, 4.9, 5.2, 5.7, 5.1, 5.8, 6, 5.8, 5.9, 5.4), Beta.Vector = c(13.6, 15.3, 15.9, 16.4, 16.1,18.5, 17.3, 18.2, 17.7, 16.4), N.Vector = c(130,

140, 150, 200, 210, 240, 300, 350, 350, 400))

# Data summary and descriptives Total number of trials: 10

# Meta-analytic results summary

R2 Trial Standard Error CI lower limit CI upper limit 0.7608 0.1577 0.4517 1.0000

R Trial Standard Error CI lower limit CI upper limit 0.8722 0.1729 0.5382 0.9695

# Obtain plot of the (trial-level) results > plot(Trial_Fit)

# Generated output:

The output shows that the treatment effect on T can be predicted with moderate accuracy based on the treatment effect on S, i.e., Rrial = 0-7608 with 95% confidence interval [0.4517; 1.000]. Notice that the confidence interval around i?2rial is wide, which could be expected given the small number of clustering units (trials) that were available for analysis.

 
<<   CONTENTS   >>

Related topics