Subscribe to DSC Newsletter

Hi to everybody,

Please does anybody can help me to solve this problem?
I have a multiple logistic regression model with more than one BINARY INDEPENDENT VARIABLES.
I need to calculate the minimum required sample size for my study, and I perfectly know how to calculate that for a model with only one binary independent variable (
But I cannot find on web any material or documentation illustrating a solution to the problem with a multiple logistic regression model with more that one binary independent variables

Will you, please, help me?????
I'd appreciate it very very much.

Views: 5307

Reply to This

Replies to This Discussion

See APPLIED LOGISTIC REGRESSION, 2nd ed. David W. Hosmer and Stanley Lemshow are the authors. Hosmer is retired from University of Massachusetts and Lemeshow last was at Ohio state. They are the experts in logistic regression.

When you say more than "one binary independent variables" are you talking about a multivariate y vector with two or more columns (0,1) responses or are you talking about a response variable with 3 or more nominal levels? Power is generally set up in based on the 'worst case' of the highest variance with the smallest true effect size. I am not sure how this would work with multivariate logits.
Thank you very much,

as for your question, I meant that I have an univariate logistic regression model (i.e., with only one dependent binary variable), where the dependent variable must be explained by a number of binary independent variables (1,0).
I have no problem when the independent variables are continuous in nature and normally distributed, because there is Hsieh (1998) who said that you can obtain the total sample size basing on the multiple correlation coefficient between Xi and the remaining predictors...
However I didn't find anything like that for the model that I talked about above.

So I hope to find in APPLIED LOGISTIC REGRESSION what I looking for.

Thank you very much!

SAS will do power for one binary predictor with optional covariates (Proc Power).

Also check this out for two binary predictors and their interaction:

Hope that helps.

I confused 'independent' with 'dependent'. I hope the book is useful. It is quite common to have only 0/1 independent variables in biostatistics. This makes the logits the log-odds.


On Data Science Central

© 2021   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service