0
$\begingroup$

I am trying to see if there is a statistical significant difference between the amount of bycatch in 5 different types of gear used by fishermen. For each gear I have varying number of replicates as there are higher number of fishermen using certains types of gears compared to others. I have converted my raw data into 'percentage of time occurred' in order to compare them. What stats test should i use? I have access to minitab, but I am a novice with it.

$\endgroup$
0
$\begingroup$

If you have at least 6 fishermen using each type of gear, you could begin with ANOVA. Since you're working with percentage data (which violates the assumptions of ANOVA), you'll want to logit- or arcsine-transform the data first.

If the ANOVA turns up significant results, you'll want to run a post-hoc test.

This is important: in the ANOVA, you'll want the identity of the individual fisherman to be a random factor.

$\endgroup$
  • $\begingroup$ unfortunately, for 2 types of gear I only have 4 fishermen in each, I thought I would just mention that after the results of the analysis. If the data wasn't in percentages but rather expressed as 0.03 (instead of 3%), would i still have to logit? ( ia am a complete novice as you can see!). I don't think my data is normally distributed either, does this affect what stats test I use?? $\endgroup$ – Guy Apr 8 '12 at 7:47
  • $\begingroup$ If the data isn't normally distributed you, could try a non-parametric test, like Kruskal-Wallis. If you went non-parametric with it, you wouldn't need to perform any data transformation on the percentages. $\endgroup$ – Julie Apr 8 '12 at 18:01
  • $\begingroup$ Done Kruskal-Wallis, and it seems to have worked. Hopefully that'll be ok! Thanks a lot! $\endgroup$ – Guy Apr 9 '12 at 20:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.