Subscribe Bookmark RSS Feed

Repeated Measures MANOVA - subject with multiple sample points of dependent variable per time period

Lyanne

New Contributor

Joined:

Sep 15, 2017

Hi,

 

I am trying to analyze data using JMP 12.01 from an experiment and I am having some difficulty.  The issue seems to be around the subject having mulitple sample points of the dependent variable per time period. 

 

Here is the experimental set-up.  Animals are placed within trays (n = 10 trays), each tray is fed one of two diets for 12 weeks.  Every three weeks five different individuals are removed and weighed but never to be put back into the tray.  We are trying to determine difference in growth due to diet over time.  So the tray is repeatedly sampled over time, but also has multiple data point per time period.  So the subject has more than one sample point per time period.

 

I have tried nesting the individual animals within the tray in a repeated measures MANOVA, but it does not let me nest:  Effects: Diet, Week, Diet*Week, animal(tray, diet), week*animal(tray,diet). The model output spits out the model: diet, week, diet*week

 

I have also tried the following nested model with the effects Diet, Time, Diet*Time, animal(tray)*week, week*animal(tray,diet), but it doesn't seem right.

 

I know I can analyze the mean growth of the animals/tray by time, but that seems like a useless waste of data and it will mask the variation in the data biasing results.

 

Can anyone help with this?  

 

 

2 REPLIES
jiancao

Staff

Joined:

Jul 7, 2014

I assume you were using MANOVA as the personality in Fit Model, correct? There is a Mixed Model personality in JMP Pro's Fit Model platform that supports nested subject effects in the analysis of repeated measures

http://www.jmp.com/support/help/13-2/Mixed_Models.shtml#

Lyanne

New Contributor

Joined:

Sep 15, 2017

Thank you.  Yes, that's right I was trying to do a Fit model with a MANOVA personality, a mixed model with a nested factor. Unfortunately, I do not have access to that version of JMP.  After some thought though I believe that the following split plot model is correct given our experimental set-up.

animal(tray, diet)&random

diet

week

week*diet