Stata Technical Bulletin
STB-58
Acknowledgment
Turnip graphs are the graph of choice in the Dartmouth Atlas of Healthcare (American Hospital Publishing Inc., Chicago,
IL 1996). The term turnip graph was coined by Jack Wennberg and colleagues at the Center for the Evaluative Clinical Sciences,
Dartmouth Medical School, because the display sometimes reminded them of turnips. Other suggested names included carrots,
flying saucers, and the Stealth Bomber.
sbe19.3 Tests for publication bias in meta-analysis: erratum
Thomas J. Steichen, RJRT, [email protected]
Abstract: This insert provides a correction to the help file for the metabias command introduced in Steichen (1998) and
modified in Steichen et al. (1998) and Steichen (2000).
Keywords: meta-analysis, publication bias, Egger, Begg.
Description
As one form of data input, metabias allows the user to provide effect estimates, theta, and standard errors, seJheta. The
published help file included a note stating that data in binary count format could be converted to the effect format used in
metabias by use of program metan (Bradburn et al. 1998). This note stated that metan automatically adds variables for theta
and seJheta to the raw dataset, naming them _ES and _seES, and that these variables could be provided to metabias using its
default input method.
This is not correct. When processing binary data, metan automatically adds variables for exp(theta) and seJheta, that is,
it is exp(theta) that is stored in variable _ES, not theta. The user must manually transform these exponentiated values back to
theta format using Stata’s log() function (or, equivalently, ln() function) before providing them to metabias. This additional
step is now documented in the help file.
Acknowledgment
I am grateful to Dr. John Moran for indirectly alerting me to this error.
References
Bradburn, M. J., J. J. Deeks, and D. G. Altman. 1998. sbe24: metan—an alternative meta-analysis command. Stata Technical Bulletin 44: 4-15.
Reprinted in Stata Technical Bulletin Reprints, vol. 8, pp. 86-100.
Steichen, T. J. 1998. sbe19: Tests for publication bias in meta-analysis. Stata Technical Bulletin 41: 9-15. Reprinted in Stata Technical Bulletin
Reprints, vol. 7, pp. 125-133.
——. 2000. sbe19.2: Updates of tests for publication bias in meta-analysis. Stata Technical Bulletin 57: 4.
Steichen, T. J., M. Egger, and J. Sterne. 1998. sbe19.1: Tests for publication bias in meta-analysis. Stata Technical Bulletin 44: 3-4. Reprinted in
Stata Technical Bulletin Reprints, vol. 8, pp. 84-85.
sbe39.1 Nonparametric trim and fill analysis of publication bias in meta-analysis: erratum
Thomas J. Steichen, RJRT, [email protected]
Abstract: This insert provides a correction to metatrim (Steichen 2000) and to its help file. metatrim implements the Duval
and Tweedie (2000) nonparametric “trim and fill” method of accounting for publication bias in meta-analysis.
Keywords: meta-analysis, publication bias, nonparametric, data augmentation.
Description
As one form of data input, metatrim allows the user to provide effect estimates, theta, and standard errors, seJheta. Both
Steichen (2000) and its accompanying help file included a note stating that data in binary count format could be converted to
the effect format used in metatrim by use of program metan (Bradburn et al. 1998). This note stated that metan automatically
adds variables for theta and seJheta to the raw dataset, naming them _ES and _seES, and that these variables could be provided
to metatrim using its default input method.
This is not correct. When processing binary data, metan automatically adds variables for exp(theta) and seJheta. That is,
it is exp(theta) that is stored in variable _ES, not theta. The user must manually transform these exponentiated values back to
theta format using Stata’s IogO function (or, equivalently, ln() function) before providing them to metatrim. This additional
step is now documented in the help file.