Hi stats whizzes!
I require some help confirming a stats method we are using. We have a behavioural task two groups of animals are learning. Basically they perform a task 40 times a day and each time the outcome is either correct or incorrect and has a time assigned to it. Based on the number correct we have set a performance criterion at which point the animals are said to have learnt the task. We would like to analyse:
1) If there is a difference in the time (days) to criterion between the two groups
2) If there is a difference in the decision time (seconds) between the who groups at various stages in learning
3) If there is a difference in the types of errors they make (which is based on comparing various indexes)
Our sample size is small (n=10, 5 per treatment). Some animals have not made the performance criterion. It has been suggested that we do a non-parametric survival analysis but:
- from my understanding our data is interval data and thus parametric
- we have done some previous stats on these animals using two sample t-tests assuming equal variances (variences tested) (i.e. we have done parametric stats)
My questions are:
1. Is it ok to mix parametric and non parametric statistics - ie, is it ok that one section is analysed parametrically and the other is analysed non-parametrically? It is likely our error indexes will be analysed with t-tests while the other two bits will be with survival analysis.
2. Is survival analysis (mantel cox wilcoxon test) appropriate for what we are asking? Our data is discrete so actuarial analysis will be used.
This task has never been performed with these animals so we have no expectations as to what distribution they fit.
Advertise on Talk Stats