Hello Annie Azmat,

If you want to explore your data after doing statistical testing against a fixed p-value, then you should be aware that repeated testing of a data set invalidates tests based on p-values. The term for this is called "data-grinding." However, if you want to explore your data for patterns, then that is legitimate, as long as you don't do hypothesis testing.

Exploring your data for patterns is called exploratory data analysis. John Tukey (as in Tukey HSD) wrote a book on the topic called Exploratory Data Analysis (1978). Have a look at the Wikipedia entry.

https://en.wikipedia.org/wiki/Exploratory_data_analysis

Best wishes, David S

Hi Annie,

From your question, it sounds like you are running an ANOVA and are asking about what followup to use if you get a significant overall effect in order to determine which pairwise effects might be significant. If that is the case, any of the procedures you mentioned as well as many more could be used. The choice depends upon how you wish to control the error rates. The choice would range from the Tukey HSD that controls the experimentwise error rate for all possible pairwise comparisons to the Least Significant Different (LSD) procedure that gives independent (individual) error rates for each comparison, but results in a highly inflated experimentwise error rate. You can find a much more detailed discussion of this in one of my publications downloadable from Researchgate titled "Control of Experimentwise and Per-experiment Type I Errors of Selected Multiple Comparison Procedures when Conducted as Protected and Unprotected Tests."

Good luck,

J. McLean

Usually, before we start performing statistical tests to endorse or confirm a research hypothesis, we need to clean the data and apply descriptive techniques like mean and standard deviation and check the data for normality.