Option for multivariate multiple analysis?I am trying to find out if there is an option in JMP to analyze data which has 6 independent variables and 3 dependent variable.
Re: Option for multivariate multiple analysis?Yes, but like any analysis, it depends on what you are wanting to do (what is the point of the analysis?). Just a few possibilities that can handle multiple factors and multiple responses: MANOVA, PLS, Neural Networks, Partition Models. There are more.
Re: Option for multivariate multiple analysis?@Dan_Obermiller Thank you for quick reply. I am working with dataset in which there independent variable is risk score assigned to 5 factors (dichotomous factors) and trying to see if these risk are related to medical care (which is defined by 3 variables out these 3, one is continuous and 2 are dichotomous. Also this is nested data so need to assign nested effect.
Re: Option for multivariate multiple analysis?I do not have the background in your area to suggest the proper analysis. The proper analysis will depend on your objectives, how the data was collected, insights to be gained, etc. This forum is not designed to tell you which analysis to do, but rather, how to use JMP to perform your chosen analysis.
A few questions/comments that may help. At the heart of all analyses will be a model. Is there a reason that you wish to have a SINGLE model to cover all three responses rather than three separate models? In many cases, there is no reason to have a single model covering multiple responses. This will be especially true in your situation since you have a continuous response and two categorical responses. The categorical responses need to be modeled differently than the continuous response. Many times, fitting a model to each response separately allows for better models than a single overall model.
The nesting MAY cause some issues as it might limit the techniques that you may wish to consider (some techniques do not allow nesting -- or are not concerned with it). I would encourage you to look at some texts on statistical modeling. Applied Linear Statistical Models by Neter, Kutner, et.al. might be a good one to consider. Your analysis will likely involve using JMP's Fit Model platform, so reading the JMP documentation on that platform could also be of use to you.
I hope this information can get you started. I wish you luck on performing your analysis and feel free to return to the forum and ask the community for help on more specific questions.
Re: Option for multivariate multiple analysis?@Dan_Obermiller I gave you all these details because I thought it will help in explaining my data. I guess my question was not clear. I am using mixed model analysis for this and right now I am using SAS (proc mixed) for this. When I tried to used mixed model personality in JMP it was not allowing me to add my dichotomous dependent variable in Y. I tried looking up in JMP documentation if I am doing something wrong but was unable to find.
I guess I'll read JMP documentation again for future analysis. Thank you though.
Re: Option for multivariate multiple analysis?JMP does not support nominal or ordinal responses in a mixed model at this time (to my knowledge, anyhow!). PROC MIXED sounds like a good option for the nominal/ordinal responses. Note that you can call a SAS program from JMP, so that may be a viable option to stay in your JMP session, but still have PROC MIXED perform the analysis on those responses.
Re: Option for multivariate multiple analysis?@Dan_Obermiller Thank you for quick response. This clears my confusion.
