0
$\begingroup$

I have a very simple sample dataset. There are four trait values from two animals (I know the sample size is very small. For each animal, one side of the body was treated with drug, and the other was not. Then trait values from both sides of the body of the two animals were measured. Body size was used as a covariate.So treatment can be paired with control in this dataset. I hope to examine the impact of treatment on trait values.

   trait   body        Treatment ID
 2217.6758 624083         N      a      
  481.8359 396332         N      b      
 3883.1055 636011         Y      a      
  777.7344 377092         Y      b 

 

m=aov(trait~body+treatment+Error(ID/(treatment),df)

However, no F or p values were calculated. Anything wrong with it?

Improve this question
$\endgroup$
3
  • 1
    $\begingroup$ Could you show your full code? The aov function usually produces the info you seek. $\endgroup$
    – Shawn Hemelstrand
    Commented Jul 1, 2023 at 11:47
  • $\begingroup$ It seems to me the body values should be the same for each related ID. That is, if ID=a is one animal, shouldn't that animal have the same body. $\endgroup$
    – Gregg H
    Commented Jul 1, 2023 at 12:22
  • $\begingroup$ Hi@GreggH, Sorry for the confusion. The 'body' value is actually a measurement of a certain body part, which is present in both sides where I measured the trait value, so they are slightly different. $\endgroup$
    – shenTTT
    Commented Jul 1, 2023 at 16:11

2 Answers 2

Reset to default
1
$\begingroup$

Your sample size for a repeated measure would be ok...if you only had the treatment factor. This would give you 1 degree of freedom (the smallest possible). However, if you include the covariate, you have no degrees of freedom left, and you will get a perfectly fitting solution (thus no error to assess statistical significance).

Improve this answer
$\endgroup$
1
  • $\begingroup$ Hi, Yes, when I remove the covariate, it did produce the usual stuff... $\endgroup$
    – shenTTT
    Commented Jul 1, 2023 at 16:08
1
$\begingroup$

Your dataset is too small for the model. The model is saturated and a perfect fit.

I added 4 lines to the data (making up numbers) and then summary(m) output the usual stuff.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.