PubBias (version 1.0)

plot_chase_observed_expected: From a meta-analysis, analyse for publication bias. Calculates observed and expected number of positive studies and P for difference.

Description

From a meta-analysis, analyse for publication bias. Calculates observed and expected number of positive studies and P for difference.

Usage

plot_chase_observed_expected(vec_r_events_control, vec_r_events_treated, vec_n_sample_size_control, vec_n_sample_size_treated, n, low.alpha, high.alpha, by.alpha)

Arguments

vec_r_events_control

an ordered vector of number of events in the untreated group of constituent studies from a meta-analysis.

vec_r_events_treated

an ordered vector of number of events in the treated group of constituent studies from a meta-analysis.

vec_n_sample_size_control

an ordered vector of the number of participants in the untreated group.

vec_n_sample_size_treated

an ordered vector of the number of participants in the treated group.

n

Number of iterations used to generate constituent study power; suggest use 10,000.

low.alpha

Lower limit of type-1 error rate used to judge whether constituent studies are positive; suggest 0.001.

high.alpha

Upper limit of type-1 error rate used to judge whether constituent studies are positive; suggest 0.3.

by.alpha

Interval of type-2 error rate at which observed and expected values and P for difference evaluated.

Value

a dataframe with columns which include alpha level, observed number of positive studies, expected number, and P for difference, OR_hat (summary measure of effect for meta-analysis) with varying levels of significance for constituent studies.

Examples

data("BMort") ## Meta-analysis of statin use (Brugts 2009, BMJ)
 Btmort<-with(BMort, plot_chase_observed_expected(r_events_control,
   r_events_treated, n_sample_size_control, n_sample_size_treated, n=10,
   low.alpha=.001, high.alpha=0.3, by.alpha=0.01))
plot(Btmort$alpha, Btmort$observed,  type="l", las=1, lwd=2, xlim=c(.0001,0.3),
   xlab=c("Significance level"),  #### Brugts study mortality outcome; n set low for speed.
   ylab=c(""), main=c("(a) Brugts; all-cause mortality."))
lines(Btmort$alpha,Btmort$observed)
lines(Btmort$alpha,Btmort$expected, lty=3)
abline(v=0.05, lty=2)
par(new=TRUE)
plot(Btmort$alpha, Btmort$p.value, type="l", xlab="",lty=4,lwd=2,
col="grey", axes=FALSE, ylab="")
abline(h=0.1, lty=2)
axis(4,las=1)
mtext(side=4,line=2.5,"P for difference")