Dear all,

I would appreciate any advice on the correct use of a Linear Mixed Model (in SPSS)

A short description of the research I’m doing:
In a group of patients (n=17), I gathered information every 15 minutes during 5 hours, giving me 20 measurements for every patient. I want to investigate whether one of my measurements influences another. The predictor is a hormone (concentration), which is known to fluctuate over the day, and the dependent is a brain scan characteristic (expressed in a number). Example of the data for one patient: (couldn't find the option to insert a table)

Pt. no.__Sampletime__Hormone__Brain scan

Firstly, I would like to use the data of all the patients to make a general assumption, but every person has their own baseline hormone concentration of course. Secondly, the brain scan data is known to fluctuate over time. So if I will find an effect of the hormone on the brain scan data, I want to make sure it’s not the inherent fluctuation over time I’m measuring.

I believe using a linear mixed model would be the solution, however I am not familiar with the details and specific settings (in SPSS).

The research I have done on LMM’s has led me to the following (when using SPSS analyze>mixed models>linear):

First window:
As subjects I take my Pt. no.
Nothing for the repeated box (and thus my repeated covariance type stays blank).

Second window:
Dependent Variable: brain scan data
Factor(s): blank
Covariate: hormone.

Third window:
Fixed effects: Add Hormone to the model. Leave all ticks as they are: ‘Build terms’ ‘Factorial’ ‘Include intercept’ ‘Sum of squares Type III’. (I am not quite sure what they mean)

Fourth window:
Random effects: I was advised to set Covariance Type to ‘Scaled Identity’. Check the ‘Include intercept’. Leave the model box blank. Subject groupings: add Pt. no. to Combinations.

I’m just putting it down word by word hoping not to miss anything.

I would be grateful for any advice and glad to give additional information.