0
$\begingroup$

hope you can help me with this issue! In my study I have 4 outcome variables which correspond to the ratings I collected for 4 different psychological dimensions (liking, comfort, approach, attractiveness). The 4 psychological dimensions are defined in the column ‘ratetype’ and the column 'ratings' reports the score. I also have 2 independent variables: 'paintingtype', which is a categorical variable with 3 experimental conditions, and ArtInterest, which is the self-rated measure on a 5-points Liker scale. This is my data structure:

Observations: 7,872
Variables: 14
$ X            		<int> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16…
$ PID                  <fct> 571b712f5d40840009c44804, 5798f7a116020100010411ac, 5…
$ ArtInterest  		<int> 4, 4, 4, 5, 4, 4, 3, 2, 3, 3, 5, 4, 4, 4, 4, 3, 5, 5,…
$ ratings              <int> 15, 72, 21, 80, 91, 13, 58, 36, 18, 70, 10, 49, 82, 1…
$ ratetype     		<fct> Liking, Liking, Liking, Liking, Liking, Liking, Likin…
$ paintingtype     <fct> Angular, Angular, Angular, Angular, Angular, Angular,…
$ paintingID    <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,…

At the moment I subsetted the data in 4 different datasets with the following code:

# # Select data for ‘LIKING’
Liking <- subset(mdata, ratetype == "Liking",
                   select=c("PID", “ArtInterest", "ratings",
                            "paintingtype", "paintingID" ))

This is the data structure of the subsetted dataset:

$ PID          <fct> 571b712f5d40840009c44804, 5798f7a116020100010411…
$ ArtInterest  <int> 4, 4, 4, 5, 4, 4, 3, 2, 3, 3, 5, 4, 4, 4, 4, 3, …
$ ratings      <int> 15, 72, 21, 80, 91, 13, 58, 36, 18, 70, 10, 49, …
$ paintingtype <fct> Angular, Angular, Angular, Angular, Angular, Ang…
$ paintingID   <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, …

and then run 4 different analysis with the model below:

m1 <-  lme(ratings ~ paintingtype:ArtInterest, random= (~1|PID/paintingID), data=Liking, method = "ML")

Q1: Would be better to run one single MANOVA instead of 4 different lme models? Q2: As every participant ('PID') rated 16 items ('paintingID') for every experimental condition, how can I control for the random effect of those two variables in the MANOVA?

Thank you in advance for your help!

EDIT: Following the suggestions in the comments, I built the following model:

m <- lmer(ratings ~ paintingtype + ArtInterest + ratetype + (ratetype | PID/paintingID) , data=mdata)

however I receive the following warning message:

singular fit

What does it mean? Should I worry about it? Any suggestion will be much appreciated!

$\endgroup$
8
  • 1
    $\begingroup$ You can always run a manova as a multilevel model. I don't quite understand your model -wnat are the 4 datasets? Perhaps include your subset command. $\endgroup$ Jul 31, 2019 at 15:59
  • $\begingroup$ @JeremyMiles thanks for your suggestion. I updated the question as you suggested $\endgroup$
    – Nicole
    Aug 1, 2019 at 9:47
  • $\begingroup$ @JeremyMiles I am not sure how to run a MANOVA as multilevel model: are you suggesting to add 'ratetype' as a fixed effect in my model? $\endgroup$
    – Nicole
    Aug 1, 2019 at 9:55
  • $\begingroup$ Here's an article about it. link.springer.com/chapter/10.1007/978-3-319-20585-4_16 . I'm fairly sure that the book Serious Stats by Baguley covers it (as do many others). $\endgroup$ Aug 2, 2019 at 16:30
  • $\begingroup$ @JeremyMiles thanks $\endgroup$
    – Nicole
    Aug 3, 2019 at 10:58

1 Answer 1

1
$\begingroup$

A short answer to the updated question: your model is too complex. You will need to either simplify your model (typically one removes more complex random effects) or to switch to a Bayesian approach to fitting your model. A more thorough explanation can be found here.

$\endgroup$
2
  • $\begingroup$ thanks for the link. I am considering running a simple repeated-measures ANOVA using the 'aov' function in R, which might be enough to model my data. $\endgroup$
    – Nicole
    Aug 6, 2019 at 10:37
  • $\begingroup$ @Nicole You can start by removing the random slopes, and then simplify it further as needed. $\endgroup$
    – mkt
    Aug 6, 2019 at 10:38

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.