multivariate analysis with categorical and continuous variables

[#1]

Hello,

I have mixed model, where I evaluated on univariant analysis by either Pearson for categorical variables or logistic regression for continuous variables significance to a binary outcome (eg mortality).

Putting those variables (continuous and categorical) into a multivariate analysis shows me a certain effect, using "Fit Model" and "Nominal Logistic" option with subsequent manual backward elimination until all p<0.05.

Now, some statistician friends told me, I cannot use categorical and continous variables in the same multivariate analysis, and should rather categorize my continuous variables into categorical variables (eg split by interquartile ranges) and use those, only categorical variables in the multivariate regression. Or vice versa, I should exclude my categorical variables in the model and use only continuous variables.

What do people think ? - Any advice much appreciated. Many thanks, Marc

[#2 · Accepted solution]

If I understand your original post correctly you are attempting to fit a nominal logistic regression model with a categorical response (binary) and both continuous and categorical predictors? If that's the case there is no reason why you can't use JMP to fit that style of model. As well as many other modeling methods ranging from neural networks to the Partition platform in JMP as well. Here's a quick example using the diabetes.jmp data table supplied in the JMP Sample Data Directory and a link to the JMP online documentation.

https://www.jmp.com/support/help/14-2/launch-the-nominal-and-ordinal-logistic-personal.shtml