Loglinear vs. Logistic Regression

[Mastani]

Hello!

I have a question about SPSS. In my thesis, I have four categorical (2 levels) variables. Three of them are IV and one of them is DV. I know I need to use Loglinear analysis or Binary Logistic regression. But what is the difference between Loglinear analysis and Logistic regression? They are both assessing the model so can I interpret the effect from the log-linear analysis? Or is it log-linear for just to show correlation? Thank you so much for your help in advance.

 

[noetsi]

I am not sure of the difference between the two, I don't do loglinear analysis, but it would certainly be correct to do binary logistic regression as compared to say linear regression if you DV has two levels.

Having said that I will point out that some economist disagree that logistic regression is necessary in many cases with binary DV (thus the linear probability model).

 

[Mastani]

I am not sure of the difference between the two, I don't do loglinear analysis, but it would certainly be correct to do binary logistic regression as compared to say linear regression if you DV has two levels.

Having said that I will point out that some economist disagree that logistic regression is necessary in many cases with binary DV (thus the linear probability model).

Thank you so much for your answer I decided to do binary logistic regression.

 

[noetsi]

I suggest, if you are new to it, that you look at something like the Sage monograph on it. https://methods.sagepub.com/book/logistic-regression
I found Paul Allison's book useful as well (its geared to SAS but still helpful). Odds Ratios are usually what you want to report not the pretty useless slopes that are essentially impossible to interpret in any meaningful way [except I guess if the slope is positive and the Wald test of significance].

 

[Mastani]

Oh really, thank you that was very helpful. I will look at these sources

 

[liweitang]

Loglinear model is only used in the circumstance with the continued dependent and independent variable. it has the elasticity interpretation and this based model is popular in empirical analysis in economics.

 

[noetsi]

Some interpretations of logistic regression, although not all, assume that there is a latent dependent variable that is interval behind the model it just gets measured as a binary.