Please Help : Multivariate Logistic Regression using JMP
Aug 27, 2009 7:31 AM(1378 views)
I am having problems understanding this procedure with JMP. I read a lot of journals articles that run multivariate logistic regression with JMP however when I look through the help section and online there is no mention of how to exactly perform this procedure. The only mentioned techniques are logisitic regression and multivariate analysis. This being said I need to run multivariate logisitic regression (I have one dependent and 3 or 4 independent variables). If someone would be so kind to answer these questions for me I would greatly appreciate it.
My Test: I have "Load Failures (ex 46.2, 89.2, 10.2, 40) " as my dependent and I have material "thickness, porosity, stiffness value, density" as my independent variable. I want to model and see if I could design a material for a specific load failure based on my independent variables.
1. Can JMP 7 and 8 run "multivariate logistic regression" analysis?
2. How do I run this in JMP? Can someone please take me through the procedure. This is such a powerful software and has a lot of choices and variables that could dramatically change the results.
This is what I have tried so far and I am confused if what I ran might be multivariate logistic regression:
logistic regression : Fit Y by X on each independent parameter.
Fit Model : Fit Y with multiple X's...this is where I am confused.
Fit Y by X and Fit model if I choose my Y variable to be nominal instead of ordinal I get an error that tells me the program cannot converge "step halving limit" etc.
So I am forced to choose "ordinal" for logistic regression and I am not very familiar with the difference this will have on my results.
Also, when I add my independent variables I could: A. Just press add B. Cross or C. Nest. This will also change my results dramatically.
If this is the correct way then how to I go about taking 3 or 4 independent parameters and testing them (together) against my load failure?
What is the proper procedure for running multivariate logistic regression using JMP. As I am a newbie to this software I would appreciate it if you could take me through it step by step.
I'm not skilled with JMP, so I'll leave the syntax questions to others.
However, you have run into some sloppy semantics in the journals. In the statistician's lexicon, "multivariate" is used for multiple dependent variables whereas "multivariable" is used for a single dependent variable. Either can have multiple independent variables. What you describe in your first paragraph looks like multivariable logistic regression. JMP can certainly do that for a binary dependent variable. Check these notes for a further discussion http://www.uncp.edu/home/frederick/DSC510/JMPlogreg.htm
If you are in JMP and go to the Help menu at the top tool bar and select indexes then statistics, you can scroll down the list and select multiple regression and the launch. It will pop-up a box for you to fill in your variables. Hope this helps...