The peer review process is the main mechanism through which scientific communities

decide whether a research paper should be published in academic journals.5 By

having other scientists evaluate research findings, academic journals hope to maintain

the quality of their published articles. However, some have warned that the peer review

process may yield undesirable consequences. In particular, the process may result

in publication bias wherein research papers with statistically significant results are

more likely to be published. To make matters worse, by being aware of such a bias

in the publication process, researchers may be more likely to report findings that are

statistically significant and ignore others. This is called file drawer bias.

In this exercise, we will explore these potential problems using data on a subset

of experimental studies that were funded by the Time-Sharing Experiments in the

Social Sciences (TESS) program. This program is sponsored by the National Science

Foundation (NSF). The data set necessary for this exercise can be found in the CSV

files filedrawer.csv and published.csv. The filedrawer.csv file contains

information about 221 research projects funded by the TESS program. However, not

all of those projects produced a published article. The published.csv file contains

information about 53 published journal articles based on TESS projects. This data set

records the number of experimental conditions and outcomes and how many of them

are actually reported in the published article. Tables 7.5 and 7.6 present the names and

descriptions of the variables from these data sets.

Question: We next examine whether there exists any difference in the publication rate of

projects with strong versus weak results as well as with strong versus null results.

To do so, first create a variable that takes the value of 1 if a paper was published

and 0 if it was not published. Then, perform two-tailed tests of the difference in

the publication rates for the aforementioned comparisons of groups, using 95%

as the significance level. Briefly comment on your findings.

Variable Description

id study identifier

DV publication status

IV statistical significance of the main findings

max.h H-index (highest among authors)

journal discipline of the journal for published articles

Attached is the dataset

This is to be completed in R