BookmarkSubscribeRSS Feed
Lder
Calcite | Level 5

A reviewer has questioned my use of Proc Mixed for repeated measures on ordinal data, saying that I should be using non-parametric methods with ordinal data. What is your advice?

6 REPLIES 6
PGStats
Opal | Level 21

Proc mixed assumes Gaussian random effects and errors.

PG
Lder
Calcite | Level 5

That is too concise a response for me to understand whether you agree with the reviewer or not. Is there a different procedure that I should use?

 

I have a continuous dependent variable (body temperature) and ordinal independent variable (comfort rating) which I measure at regular intervals over a period of time. If proc mixed is not appropriate, then please advise what would be?

Lder
Calcite | Level 5

This is the script - Sens is ordinal (a rating scale) where bodytemp is continuous.

 

proc mixed data=mydata method=ml covtest asycov ;
class subject_ID ;
model sens = bodytemp /DDFM=KR CL s noint;
repeated / type=UN subject= subject_ID R;
run;

StatDave
SAS Super FREQ

For an ordinal response, you should use either PROC GLIMMIX or PROC GENMOD instead. 

sld
Rhodochrosite | Level 12 sld
Rhodochrosite | Level 12

I agree, if in fact SENS is intended to be the response variable. That is not yet clear to me, given the contradictory posts.

sld
Rhodochrosite | Level 12 sld
Rhodochrosite | Level 12

The "dependent" variable is the response variable. In one post you identify it as body temperature. It is not unreasonable to assume that body temperature follows a normal distribution, as a first pass.

 

The "independent" variable is the explanatory or predictor variable. You identify it as comfort rating, and you say that comfort rating is measured on an ordinal scale. You might choose to treat comfort rating as if it is nominal and use an ANOVA-like model. Or you might choose to treat comfort rating as if it is continuous and use a regression-like model. In my opinion, it depends on the shape of the relationship between the response (dependent) variable and the explanatory (dependent) variable as well as the number of levels of the explanatory variable.

 

BUT!!! in a subsequent post, you reverse the identities of response and explanatory: in your MIXED code, you identify SENS as the response (dependent) and BODYTEMP as the explanatory (independent) variable. So, I have no idea what role your variables play in your study, or how to express those roles in a statistical model. 

 

Plus in your model, you use the REPEATED statement, but you have not identified a repeated measures factor in either the MODEL or the REPEATED statement.

 

To get appropriate advice, you need to provide both clarity and more information about your experimental design. In addition, I highly recommend studying SAS® for Mixed Models.

 

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

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
  • 6 replies
  • 2457 views
  • 0 likes
  • 4 in conversation