im conducting a study on the characteristics of tb patients in terms of their treatment outcome (successful and unsuccessful). my independent variable/data sets include sociodemographic/economic factors(age, gender, income, civil status etc) of tb patients. is it possible to use cluster analysis out of this? can cluster analysis perform the similarities of patients who have successful outcome and patients with unsuccessful outcome base on the dependent variables? i'm confused between cluster analysis and discriminant analysis. please help. thanks

Cluster analysis determines how many ‘natural’ groups there are in the sample. It also allows you to determine who in your sample belongs to which group.

Discriminant analysis uses a collection of interval variables to predict a categorical variable (in your case successful versus unsuccessful outcome).

Which you want depends on exactly what you want to ask. If you want to see which variables predict outcome, and how much of the variance in outcome is predicted by your variables than you need discriminant analysis. If you want to see how patients group or cluster together on independent variables then you need cluster analysis.

I found this chapter helpful on cluster analysis http://www.uk.sagepub.com/burns/website%20material/Chapter%2023%20-%20Cluster%20Analysis.pdf

They also have one on discriminant analysis http://www.uk.sagepub.com/burns/website%20material/Chapter%2025%20-%20Discriminant%20Analysis.pdf

links didn't post well,but if you cut and paste they should hopefully work

Hi,

maybe this is completely irrelevant, but if you have participants that they are dependent (like patients clustered in hospitals or in neighborhoods), then you might use multilevel logistic regression ..
Just a suggestion :) Then you can find Intra-class correlation, odds ratio and CI%

so in using cluster analysis, one could not predict the outcome base on groups/cluster? my research problem is to determine the similarities and characteristics of MDRTB patients in terms of treatment outcomes. i wonder how can i utilize cluster analysis to answer this problem.