Multivariate analysis for different types of variable using SPSS?

Dear all,

Thank you all for answering my previous questions on my project. My boss just gave me the green light to start writing a manuscript draft and I thought I might want to add a few more analyses, specifically the multivariate analysis. I watched a few tutorial videos on YouTube to understand the basics but I am still having a bit trouble on what I want to accomplish:

1. Most of the videos on YouTube only discussed continuous variables to compare means. Many of the variables in my data table were continuous but some were categorical (i.e. are you a smoker or not), while others are discrete (severity of disease stage 1, 2 or 3). Is there a multivariate analysis that allows me to combine all these variable together and ask the following question: which variables are statistically different?

2. The key figure in the manuscript includes several Kaplan-Meier survival curves that compare procedure patency between genders, followed by log-rank to compare difference. Again, given all the variables collected in my table for each patient, is there a multivariate analysis to compare the log-rank difference after taking all variables into account?

My stats and programing knowledge are rudimentary at best so I can only handle the graphical user interface in SPSS. Coding with SAS is a bit beyond my capability. Thank you in advance for your time and help.



Less is more. Stay pure. Stay poor.
@VascularGuy - you need to move over to proportional hazards models (e..g, SAS - Proc Phreg). This procedure allow you to use multiple variables and control for them at the same time. You will end up with hazard ratios as your estimand. It is usually a little tricky to directly compare multiple variable to each other. Key reason may be, one variable has a larger coefficient (estimate), but also is less precise (wider standard error - confidence interval) compared to another variable with a slightly smaller coefficient but is more precise. You can report variable values and leave it up to the reader to discern the rankings of variables - if there is any close calls.

Also, I am imagining you are writing about multiple regression not multivariate regression. The former has multiple predictors and the latter has multiple outcomes.