In considering doing a logistic regression using the Enter method, it was suggested that I may want to consider doing a sequential LSR or stepwise logistic regression instead.
What exactly are the advantages and disadvantages of each of these options, and what criteria should I use in order to decide which method is best?
What are the conditions under which logistic regression is more preferable compared to ordinary least squares regression? Is there a logistic regression equivalent to the R square statistic?
* Into: The goal in multiple regression (linear or logistic) is to measure the strength of predictors on an outcome, i.e., the numerical effect independent variables have on the dependent variable(s). Logistic regression is a sequential method used for categorical data when the outcome is dichotomous, usually coded as 1 or 0. The input variables for logistic regression can be a mix of continuous, interval, and nominal data and need not be normally distributed. Statistical software like SPSS is used for conducting logistic regression.
In considering doing a logistic regression using the Enter method, it was suggested that I may want to consider doing a sequential or stepwise logistic regreession instead.
Question 1: What exactly are the advantages and disadvantages of each of these options, and what criteria should I use in order to decide which method is best?
* Answer 1: Ultimately, the choice to use sequential (vs standard) least squares regression (LSR) or logistic regression is whether or not you have a continuous or dichotomous type independent variable. Sequential LSR uses the same method as standard LSR, but with deliberate discretion on the part of the researcher to 'try' various combinations of variable interactions 'sequentially'. In sequential LSR analysis, if you plug in all independent variables in the first step of the sequence, you are really doing standard LSR, and you will get identical results. Stepwise logistic regression is considered by some researchers as sort of a short-cut when applied to data that should be analyzed with ...
In 926 words, the solution provides an overview of sequential least squares regression (linear) and logistic regression and then describes some of the advantages and disadvantages of each model. The solution then zeros in on the specific circumstances when logistic regression is a better choice. Logistic regression best-practices are included as is a short discussion of the measures of fit and reliability for logistic regression (R-square equivalents). Suggestions for further reading and a PowerPoint slide deck are also provided.