Go back to DSc8450 home


Empirical Issues in The Decision Process For Multivariate Analysis

Stage 1: Define The Research Problem, Objectives and Multivariate to be Used

Is the problem suitable for multivariate analysis?
Can specific measures be identified for the concepts of interest?
Which multivariate technique is best suited to the research problem?

Stage 2: Develop an Analysis Plan

How does sample size affect your results?
Are the variables of the correct measurement type? If not, can they be transformed?
Can nonlinear relationships be identified and represented in the variate?

Stage 3: Evaluate the Assumptions of the Multivariate Technique

Have the missing data characteristics of the data been assessed?
Are there any outliers that might affect the results?
Have the underlying assumptions been tested empirically?

Stage 4: Estimate the Multivariate Model and Assess Overall Model Fit

What is the statistical power of the multivariate technique?
What are the measures of overall model fit and how are they interpreted?
How do you interpret the errors of prediction or explanation?
What empirical bases are available for possibly respecifying the multivariate model?

Stage 5: Interpret the Variate

Are the results evaluated with some measure of statistical significance?
What results come from evaluating the variate versus evaluating individual variable(s)?
How do you compare the impact of different variables on the results?

Stage 6: Validate the Model

Can you form estimation and holdout samples from the original sample?
Can you use techniques which validate through the omission of single cases, such as bootstrapping?
How do you compare and evaluate the differing results obtained in your validation efforts?

Go back to DSc8450 home