# Is there any good reference which statistical test should be used given study design?

I am looking for a source (infographics, picture, graph, list, book, algorithm or whatever) that clarifies which statistical test should be used when.

I mean what do I use when I have two unpaired groups and want to compare means of continuous numerical data?

What test do I use when I have two groups and want to compare ordinal data?

And so on....

Any good advice?

## 2 Answers

What test you should use (in SPSS) depends on more factors then than just the 'scaletypes'of your variables.. If I understand correctly from you description your independent (predictor) variable is dichotome data (= 2level form of categorical), and your dep var (outcome) is ordinal (= also form of categorical). Technically (if K-S'test en Levene's are non-significant) you should use Chi-Square test. (under descriptive > crosstabs).

(However if your dep var is a 5 or more likert scale you are likely to have more power (without breaking rules) with a simple paires sample t-Test..)

Good luck!

• My question was general and is not about my current study. Things that determine the test to be used are: distribution of the variable we want to test, paired or unpaired samples, type of scale of the variable we are measuring (but this is probably covered by the distribution), number of groups (two or more than two), what did I missed? – WojciechF Sep 18 '14 at 20:19
• AH I see, you missed just one I believe. The amount of predictor (=indep. vars) variables (this is different from 'number of groups' as things to determine the test to be use..) should be added. – Steven B. Peutz Sep 22 '14 at 10:13

This might be helpful to all in doubt which test should they use. Best reference I found so far: http://www.ats.ucla.edu/stat/mult_pkg/whatstat/default.htm