9.1 What To Do When

This section is help to determine what test to perform when. It is only to be used as a rough guideline and is accurate most of the time but not always. In particular, the case of a small sample this section could be misleading. This section will aid greatly in understanding why and when a statistician may perform certain statistical tests. It is the opinion of the author that often it is important to consult a statistician and that most if not all introductory courses are not enough for a researcher to lead the statistical aspects of the project.

What to do when is mainly determined by the number of variables and the type of variables you are working with. Yes, it is often this simple, for the general situation. It is necessary to check assumptions but below is a starting point.

First off, what is the number of variables and the type of variable(s):

1. One variable: categorical
2. One variable: continuous
3. Two variables: categorical and categorical
4. Two variables: categorical and continuous
5. Two variables: continuous and continuous
6. Multiple variables: Continuous dependent
7. Multiple variables: Categorical dependent (beyond the scope of this text)
8. Exception: Time series data - use time series techniques

1. Categorical
   1. Two Categories
      • Binomial Test or Z approximation for test of proportions
   2. More than two categories
      • Chi-square test
2. Continuous
   • t-test

1. Categorical and Categorical
   1. Two Categories For Each Variable
      1. If one or two sided test: Two Sample Z-test of proportions
      2. If two sided test: Chi-Square test
   2. More than two categories for at least one of the variables.
      • Chi-square test
2. Categorical and Continuous
   1. Two Categories
      • Two sample t-test
      • Paired t-test (if data is paired)
   2. More than two Categories
      • ANOVA
3. Continuous and Continuous
   • Correlation and simple linear regression

1. Multiple variables: Continuous dependent
   • General linear model
2. Multiple variables: Categorical dependent (beyond the scope of this text)
   1. Two Categories
      • Logistic regression
   2. More than two Categories
      • Discriminant analysis

• Time series data - use time series techniques