# Conducting several simple logistic regressions vs conducting one multiple logistic regression

I am conducting analysis on my honors thesis. I have one binary outcome variable, having high or low intimate relationship quality. I want to use logistic regression to predict the odds of having high or low intimate relationship quality. I have 5 predictor variables: depression, anxiety, illness acceptance, self-concept and physical self-concept. I have been advised not to conduct one multiple logistic regression with all the predictors included, but instead conduct 5 simple logistic regressions for each individual predictor separately. Is this an appropriate approach or should I conduct a multiple logistic regression with all 5 predictors in the one model?

EDIT


Advised by my supervisor, they said that was how I should conduct the analysis but didn't give many reasons. My sample size is 115 with 33 in low intimate relationship quality and 82 in high intimate relationship quality. Data was collected via online surveys. Basically, I am unsure if conducting 5 separate analyses is the best way to conduct my analysis, it seems that when I do 5 separate logistic regressions all predictors are significant, but if I conduct one multiple logistic regression with all predictors, only one predictor is significant.

• Adviced by whom? What was the reason given? What is your sample size? How was data collected? We need more information ... – kjetil b halvorsen Sep 10 '15 at 8:58
• Advised by my supervisor, they said that was how I should conduct the analysis but didn't give many reasons. My sample size is 115 with 33 in low intimate relationship quality and 82 in high intimate relationship quality. Data was collected via online surveys. Basically, I am unsure if conducting 5 separate analyses is the best way to conduct my analysis, it seems that when I do 5 separate logistic regressions all predictors are significant, but if I conduct one multiple logistic regression with all predictors, only one predictor is significant. – Tom Sep 10 '15 at 9:03
• You can present both of them, and try to understand which aspects of the data can explain that difference. Is the correlations between the predictors high? Are the five constructs numerical variables, or categorical? And, please, answers to questions for more info should be as an edit to the original post, not in the comments! – kjetil b halvorsen Sep 10 '15 at 9:35
• Is your supervisor a statistician? If not you should ignore his advice and correct him. I've heard about supervisors telling students to use linear regression for time series forecasting... – Digio Mar 28 '19 at 22:47

## 1 Answer

That advice from your supervisor seems strange to me, the usual approach would be to use all predictors in one model. You can find more discussion in a good answer: Choosing the proper statistical approach for glm

But it is also useful, as a way of understanding the data better, to fit multiple models, including also models with only one predictor. Here is a short list of post about logistic regression strategies.