topic Modelling multilevel data in Statistical Procedures

Modelling multilevel data

Melk — Wed, 06 Nov 2019 23:39:53 GMT

I have a continuous outcome measure. The measurement was taken under 3 different conditions, in 2 locations on the body for each patient in the study. We are looking at the measurement also in 2 different ways: and average across all the "readers" and then also 1 "skilled" reader. Our research question is, is the average measurement taken from all readers as good as distinguishing between the 3 different conditions as the 1 skilled reader? Also, are the measurements taken from the 2 locations significantly different? I am not able to figure out how to write my mixed model for my random effects and fixed effects to answer these questions and would be very appreciative of any suggestions.

I suppose the data looks something like this:

ID Y Condition Location Reader

1 A A 1

1 B A 1

1 C A 1

1 A B 1

1 B B 1

1 C B 1

1 A A 2

1 B A 2

1 C A 2

1 A B 2

1 B B 2

1 C B 2

Re: Modelling multilevel data

PaigeMiller — Wed, 06 Nov 2019 23:45:14 GMT

I don't see a continuous outcome (response) measure.

The statement "Our research question is, is the average measurement taken from all readers as good as distinguishing between the 3 different conditions as the 1 skilled reader?" kind of makes this continuous variable a PREDICTOR rather than a RESPONSE.

Can you make explicitly clear which variables are predictors, and which variables are responses? Or is there no such distinction in this data?

Can you state the hypotheses of interest in mathematical terms? Such as: mean response of continuous variable of Reader 1 = mean response of continuous variable of Reader 2.

Re: Modelling multilevel data

Melk — Wed, 06 Nov 2019 23:56:19 GMT

The response variable is Y.

I am am having a hard time translating this myself to a workable hypothesis. I want to know I suppose whether the mean value of Y significantly differs by condition, by reader(1 is the skilled, 2 is the averaged), and by location? So I am interseted in their fixed effects, although they are all nested effects at the same time.

Re: Modelling multilevel data

ballardw — Thu, 07 Nov 2019 00:28:03 GMT

@Melk wrote:

The response variable is Y.

I am am having a hard time translating this myself to a workable hypothesis. I want to know I suppose whether the mean value of Y significantly differs by condition, by reader(1 is the skilled, 2 is the averaged), and by location? So I am interseted in their fixed effects, although they are all nested effects at the same time.

The way your example data is presented it appears that there are no values for Y. Is that the actual case?

Re: Modelling multilevel data

Melk — Thu, 07 Nov 2019 00:30:10 GMT

No, I have outcome data, I just didnt write any dummy data in the example as I didn't think it would provide any additional relevant information making up values having already written already it was a continuous variable.