BookmarkSubscribeRSS Feed
Migueldoisa
Calcite | Level 5

Hello,
I would like to ask for a clarification. If anyone could help me, I will be very much obliged. I have an experiment with five treatments (diets fed to animals in different parks), the response score was measured everyday, for 35 days. response variable is ordinal with three levels: 0 no effect, 1 mild effect, 2 high effect.

 

I have tried the following approach

proc glimmix data=score method=?;
class diet score;
model score = diet / link=cumlogit dist=multinomial;
random time / sub=park type=ar(1);

 

I have tried several 'method=' but would like advice which would best apply.

 

Also, in this paper with a similar approach, authors calculate the probability of score HIGH, which I would also like to apply here. Authors state the following formula but I'm not aware how to obtain the logit values in my case:

p = exp(logit)/(1 + exp(logit))

https://www.mdpi.com/2076-2615/13/21/3350

 

Any ideas appreciated.

1 REPLY 1
Migueldoisa
Calcite | Level 5

For anyone who might see this,

I found this very nice tutorial on the subject

https://agstats.ca/2019/06/25/working-with-binary-and-multinomial-data-in-glimmix/

and am currently working with following script:

 

proc glimmix data=score method=laplace;
class diet;
model score = diet / link=cumlogit dist=multinomial oddsratio(diff=all);
random time / sub=park type=ar(1);
 
contrast "Control vs others" diet 4 -1 -1 -1 -1;
estimate "score 0,1 in Control" intercept 0 1 0 diet 1 0 0 0 0 /ilink;
 
I have removed score from class statement as it doesn't seem to make a difference.
I have added oddsratio although I don't think I'll need it seeing as I'm using contrasts?
And from that tutorial I also found I can use estimate to compute the probability of score 2, which is what I need. From what I understand, for diet 1, the statement "intercept 0 1 0 diet 1 0 0 0 0" would compute the cumulative probability of score 0 and 1. Thus, the probability of score 2 would be 1 minus that. Is this correct?
However, I'm getting a warning: 
"WARNING: More coefficients than levels specified for effect Intercept. Some coefficients will be ignored."
 
I have 3 coefficients, seeing as I have 3 levels of score (0, 1, 2). Am I misjudging something?
 
I still don't know how to determine/choose which method= statement to use.
 
Attentively,
 

sas-innovate-2024.png

Available on demand!

Missed SAS Innovate Las Vegas? Watch all the action for free! View the keynotes, general sessions and 22 breakouts on demand.

 

Register now!

What is ANOVA?

ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 1 reply
  • 405 views
  • 0 likes
  • 1 in conversation