Choose Language Hide Translation Bar

Re: How to analyze unbalanced data which three of independent factors are unbalanced levels?

Hi @Shad ,

 

How about trying Stepwise regression with forward selection?  That way you will only include interactions that have data. Unbalanced is OK, but if there is missing data for a factor in an interaction, then you cannot use it. there is nothing one can estimate because the data is not there. If you have too many model effects and not enough data, then you cannot estimate all of the effects. 

 

Stepwise with forward selection can help in these cases. Same with Decision trees.

 

Hope that helps.

 

Chris

Chris

Chris Kirchberg
Principal Systems Engineer, Life Sciences - JMP Global Technical Enablement
SAS Institute, Inc. - Denver, CO
Tel: +1-919-531-9927 ▪ Mobile: +1-303-378-7419 ▪ E-mail: chris.kirchberg@jmp.com
JMP – A Division of SAS Institute | www.jmp.com
Shad
Occasional Contributor

Re: How to analyze unbalanced data which three of independent factors are unbalanced levels?

I did the stepwise with forwardn selection and minimum BIC rule. Now it ended up with some significant factors. If I want to compare each factor level or their interaction means, how I can do that? I want to see if there is any significant differences in response among predictors levels or their interaction combinations. 

0 Kudos

Re: How to analyze unbalanced data which three of independent factors are unbalanced levels?

Hi @Shad ,

 

Once the terms are selected, there is a button that says make model. For categorical models, you can add the whole effects and then the interaction terms and remove the parameterized effects.

 

Here is a link regarding Stepwise Regression that should help.

https://www.jmp.com/support/help/14/stepwise-regression-models.shtml#

 

Best,

Chris

Chris Kirchberg
Principal Systems Engineer, Life Sciences - JMP Global Technical Enablement
SAS Institute, Inc. - Denver, CO
Tel: +1-919-531-9927 ▪ Mobile: +1-303-378-7419 ▪ E-mail: chris.kirchberg@jmp.com
JMP – A Division of SAS Institute | www.jmp.com
0 Kudos