DAG-Stat: Diagnostic & AGreement Statistics. This is a freely available Excel workbook, written by Andrew MacKinnon. It provides a comprehensive range of statistics calculable from 2x2 tables that are useful in evaluating diagnostic tests and interrater agreement.
kappa.exe - a program for calculating Cohen's kappa and weighted kappa for 2 raters. (It is stored as a zip-file because Angelfire/Lycos blocks download of .exe files.)
John Uebersax's Statistical Methods for Rater Agreement. This site is a resource for the analysis of agreement among raters, diagnostic tests, observers, judges or experts. It contains background discussion on different methods, examples, references, software, and information on recent methodological developments.
the analysis of exact probabilities (p values) as suggested by
Rosenthal
the analysis of effect sizes d, following the methods of either
Glass or of Hedges and Olkin
the analysis of effect sizes r as recommended by Hunter, Schmidt, and
Jackson
EasyMA: Free
software for meta-analysis of 2x2 tables, developed in the Department of Clinical Pharmacology of the cardiological hospital in Lyon, France. Note that I have not been able to run this DOS-based software on some newer and faster machines--but if it will run on your machine, it's pretty good. Also, the manual is very good.
My own programs. I have written 3 programs that implement the procedures described by Fleiss (1993). Zip-files are available for download.
MACONT.EXE is for continuous (or interval scale) variables, and requires one to enter (or read from a text file) a study number, N, Mean, and SD for both the Treatment and Control groups for each trial or study.
MA2x2.EXE is essentially the same, but is for use with binary outcomes. Instead of entering means and standard deviations, however, you can simply enter the 4 cell counts.
FLEISS.EXE is a general purpose program that requires you to enter a study number, total N, some measure of effect size (e.g., a logit) and its variance for each study. It requires a bit more data preparation than the others, but can be used for any measure of effect size. An SPSS syntax file version is also available.
All 3 programs store the results in a text file. I don't have routines for computing p-values in Pascal (the language the programs are written in), so these are not given. However, can be obtained easily using stand-alone programs like StaTable, which can be freely downloaded from the Cytel website. Alternatively, you could use Jerry Dallal's Statistician's Swiss Army Knife.
