0
$\begingroup$

I am struggling to choose the correct statistical test for my data. I want to know if species richness is statistically different between these three habitats. Im not interested in whether it has changed over time. Only whether, when you take account that it does vary over time, you consider it to be different or not between habitats.

I have considered repeated measures ANOVA but from what Ive read this is not applicable if you have more than one observation for each factor level (replication). Maybe MANOVA or mixed models are more applicable? but Im not certain and Ive played around with a mixed model in lme4 but wasn't sure what factor to put in the Error term. The data can be normalised and with equal variances when transformed, but I have a very unbalanced design. So for each of those error bars the data are from 2-12 sites (2 sites for mudflat, 2 for Fringe, and 12 sites for Mangrove).

Species richness in habitats over time

Any help would be greatly appreciated!!

$\endgroup$
  • $\begingroup$ hello Which code did you use for this graph i need to get like this graph could you please the code $\endgroup$ – ugur Emre Nov 26 at 5:09
0
$\begingroup$

Repeated measures omnibus is applicable. If you don't care about time, then take the average over replications so you end up with just site x habitat. If you think that the time variable matters, things will get more complicated because 'time' isn't a fixed factor - the time ranges between replications aren't equivalent. If you had to do that, I suppose you could somehow add actual time as a covariate to somewhat detrend the data.

$\endgroup$
  • $\begingroup$ By Omnibus do you mean ANOVA? Its not that I dont care about time sorry. I just dont care if the species richness is different at the beginning of the time series compared to the end. I do care about it if it is causing variation in my data that should be accounted for in a statistical test (which I think it is). I thought the problem withe using RM-ANOVA was that my study is unbalanced? Maybe that's why you're suggesting taking a mean but then aren't I losing a lot of information? Is a mixed model not suited to address this? $\endgroup$ – Guidofish Dec 2 '17 at 0:24

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.