Hello everybody,
I've collected some metrics from a group of subjects on two different days (paired). Now I want to check if there is a statistical difference in one of the measures (i.e Systolic Blood Pressure), but I also want to use another variable as a covariate (also collected in both days; i.e Stress level - categorical data ranging from 1-10).
Can everybody help me with an example R code example and/or some explanation to perform this test?
Thanks in advance,
Rhenan Bartels
Attached is a file with some sample R code.
Someone may have a different or better model.
My example creates a mixed-effects model using the lme4 package. Essentially, it is just a model with SBP as the dependent variable, and Time and SL as independent variables, plus Individual is treated as a random variable, giving each Individual their own intercept in the model. This has the effect of pairing the T1 and T2 values for each individual, as if each individual is a block.
Note that the lsmeans are different than the arithmetic means. This is because the lsmeans are adjusted for the effects of SL and Individual.
The plots are only marginally helpful, because they can't show the effect of the all the variables in the model in an easy way.
The final two plots are useful for checking the residuals of the model.
Just an add-on:
If you have fully paired data (no missing values), then you can treat "individual" as a fixed factor. This would be the 1:1 equivalent to a paired test:
SBP ~ Individual + Time + Covar1 + ...
Another possibility is to calculate the differences "after-before" first and then fit the model on the differences:
deltaSBP ~ Time + Covar1 + ...
And finally a note:
As your response is a score (1-10), you might want to think about the sensibility of the normal assumption. If the mode is close to the limits you might consider a beta-regression model. If the distribution is not unimodal the entire story might be more complicated and analyzing a mean difference in scores might not be very sensible at all.
Hi Mr. Mangiafico and Mr. Wilhelm,
thank you very much for the answers, they were exactly what I was looking for!
Best wishes,
Rhenan
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA