Premium
The logistic analysis of epidemiologic prospective studies: Investigation by simulation
Author(s) -
Lilienfeld David E.,
Pyne David A.
Publication year - 1984
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.4780030104
Subject(s) - statistics , logistic regression , standard deviation , regression dilution , regression analysis , standard error , mathematics , population , statistical significance , econometrics , medicine , nonlinear regression , environmental health
We performed a Monte Carlo computer simulation of the Walker‐Duncan logistic regression technique in a typical epidemiologic prospective setting and analysed the results with respect to the accuracy and reliability of the regression estimates and the associated statistical significance tests ( Z ‐tests). The results strongly suggest that the estimates were neither accurate nor reliable. The magnitude of the difference between the average estimated regression coefficient and its true population value did not necessarily decrease as the sample size increased. The average estimated standard deviation of the estimate of the regression coefficient either overestimated or underestimated the actual standard deviation, the former occurring most, but not all, of the time. The significance tests (a two‐tailed Z ‐test with a significance level of 0.05) had actual type I errors ranging from 0.00 to 0.24 for different samples. This approach is therefore inadequate as an epidemiologic tool for analysis of a Framingham‐type prospective study. Further simulation studies are indicated.