How can I make a predictive model for a disease x by using 2 or more different laboratory test values? I have 4 groups, normal control, diseased control without a complication, early stage (asymptomatic), late stage (symptomatic and tissue damaged)
Hello Numan. I assume the lab test values are dependent (outcome) variables, is that right? Assuming they are scores for which means & SDs are suitable descriptive statistics, you could use the MIXED procedure, with occasions of measurement clustered within patients. See the SPSSX-L thread linked below for examples of how to use MIXED to estimate a multivariate model (i.e., a model with more than one outcome variable). HTH.
- http://spssx-discussion.1045642.n5.nabble.com/Multivariate-Analysis-in-MIXED-td1092023.html
First try ANOVA to test significance then try modeling if permitted with levels of your data.
Hello Numan. For SPSS, the books authored by Heck, Thomas & Tabata are quite good. See this website for more info:
- http://www2.hawaii.edu/~ltabata/mlm/index.html
For Stata, the book by Robson & Pevalin provides a nice introduction to multilevel modeling:
- https://www.stata.com/bookstore/multilevel-modeling-in-plain-language/
And of course, in Stata you can just type -help mixed- in the Command window and click on the link to the PDF file, which will have some nice examples.
HTH.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS