Which statistical test to use?

Hi guys,

I am posting to gain a bit of advice on which statistical test to use for our current research study. We are due to meet with our statistician next week but would like to prepare a little before we meet.

Our study is investigating whether differences exist in training stress between different types of training in professional sports players. E.g. is there a difference in the total distance covered during speed training compared to conditioning training.

We collected multiple measures of training stress from 17 players during individual training sessions over the course of a two year period. We have ~ 700 individual training sessions (and the associated measures of training stress) in total split between 6 different training types of training. Every individual training session (of the 700) has a complete data set for the measures of training stress.

However, due to the nature of professional sports, on a group level, teams will undertake certain training types more often than others and so we have training types with a considerably greater number of training sessions than other training types.

For example:
Training Type 1: n = 88
Training Type 2: n = 263
Training Type 3: n = 170
Training Type 4: n = 99
Training Type 5: n = 60
Training Type 6: n = 41

Also, due to the longitudinal design and the nature of professional sports (e.g. injury, illness, transfer to other clubs etc), individual player contributions towards the total number of sessions for each training type also varies considerably.

For example:
Training Type 1: Total n = 88 (Player One: n = 8, Player Two: n = 17, Player Three: n = 24 etc etc).

We only want to determine the differences in the training stress measures between the different training types on a group level.

Hopefully I have explained the design of the study in enough detail to determine the appropriate statistical test to use. Any pointers in the right direction would be much appreciated!


TS Contributor
If each of you n=17 players underwent each training type
at least once, then you could aggregate each player's
stress within each training type, e.g. using his/her intra-
individual mean or median. You could then compare the
6 types by using repeated measures ANOVA or Friedman test.

Alternatively, some more sophisticated approach could be
used (multilevel modeling or gerealized estimating equations),
especially if not every player experienced all 6 training types.

With kind regards