turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Analysis of %Growth in Mussels-How do I deal with ...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Feb 28, 2017 9:27 AM
(734 views)

I am trying to analyze the data from a field experiment I ran in 2015. I have 4 fixed factors with various levels (we'll call them a,b,c, and d). Factor c is nested in b. I have 1 random factor (blocking - randomized complete block design). My response variable is % growth in length (mm) ranging from -25% (some shrunk) to 70% and my n is 304. I used the fit-model platform to general a linear mixed model that came out pretty much as I expected - with factors a,b, c, and axb having significant effects.

My problem is that when I go to check the residuals for the assumptions of normality, the data are not normally distributed. Looking at the quantil plot reveals an s curve that just strays a little bit from the normal line. In my search I have seen these type of curves called fat tailed, and some have instructed that all %change data are actually Cauchy distributed rather than Gaussian (normal) and should not be analysed with ANOVA. That's great, but I have no idea what to do with that information. I understand how to do ANOVA and ANCOVA, but beyond that, I'm a little lost. I have tried the standard transforms (log, sqrt) with no luck.

I know that using a generalized linear mixed model allows you to work with data that are not normally distributed, but I don't have JMP pro (which I think is the only way to use glmm). Any tips or advice would be greatly appreciated. Thanks for your time.

EDIT: added number of observations (n=304)

1 REPLY

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Mar 3, 2017 6:23 AM
(676 views)

Hi -- @Nathan_Haag

Neither JMP 13 nor JMP Pro 13 support generalized linear mixed-effects models. We have this feature-request documented so development is aware of the interest.

Thank you,

Dan

Check out the JMP blog: jmp.com/blog

This widget could not be displayed.