SEDL / STP / STP1007-EB / STP10298S

Comparison of Several Computer Programs for Probit Analysis of Dose-Related Mortality Data

Roberts, MH
Professor of marine science, Virginia Institute of Marine Science, School of Marine Science, College of William and Mary, Gloucester Point, VA

Pages: 13    Published: Jan 1988

Download this paper for $25 PDF (196K)          View License Agreement
Abstract

Five readily available computer programs for probit analysis were evaluated for input/output options and reliability of output: (1) DULUTH-TOX from Charles Stephan of the Environmental Protection Agency's Environmental Research Laboratory, Duluth; (2) ASTM-PROBIT from the ASTM Guide for Probit Analysis, Draft 2; (3) UG-PROBIT, which was written by statisticians at the University of Guelph, Canada; and (4) SPSS; and (5) SAS. SPSS and SAS are probit procedures available in standard statistical packages.

Except for UG-PROBIT, the programs yielded essentially identical median lethal concentration (LC⁵⁰) values for the 20 data sets tested. DULUTH-TOX and ASTM-PROBIT include an objective evaluation of the validity of input data and calculated results, but they lack graphical output. SAS-PROBIT provides graphical output superior to that from SPSS-PROBIT. SPSS-PROBIT yields inappropriate fiducial limits in some cases.

UG-PROBIT, SPSS-PROBIT, and SAS-PROBIT purport to consider control or "natural" responses. Only SPSS-PROBIT and SAS-PROBIT actually adjusted observed treatment responses for the control response.

SPSS-PROBIT and SAS-PROBIT are currently available only for mini and mainframe computers. All other programs can be implemented on microcomputers as well. DULUTH-TOX and ASTM-PROBIT, in all implementations, have a major advantage in that objective tests are included to determine the validity of input data and to guide the interpretation of output. The commercial statistical programs have the advantages of graphical output and a method for handling control mortality.