Hi all,
I have data from 25 subjects, who are divided into 3 groups (healthy subjects, mild patients and severe patients). Each subject performed an inspiratory loading test that consisted in breathing against 6 different loads (requiring increasing effort) of 0, 12, 24, 36, 48 and 60 % of subject's maximum pressure. Several continuous outcome measures were taken from each subject at the different loads. All outcome measures are expected either to increase o decrease linearly with increasing load. So, I have a between-subject factor (group) and a within-subject (repeated measures) factor (load). I want to implement a mixed effect model for each outcome measure. Since different subjects have different starting values of the outcome variable and different increasing/decreasing rates (slopes) over loads, I have defined a mixed model with random intercepts and random slopes in R using the lme function (nlme package), as follows:
outcome measure ~ load * group + ( 1 + load | subject)
I don't know whether use load as a categorical factor (levels L0, L1, L2, L3, L4 and L5) or as a continuous variable (0, 12, 24, 36, 48 and 60). In my study it is important to check for between-group differences at the different loads, so I my first option is using load as a factor. However, I am unsure between these 3 options:
1) Using load as a factor. In this case, can I use it in the random part of the model for random slopes? I mean, does it make sense to define a random effect for a factor?
2) Using load as both a factor and a continuous variable. Could I use load as a factor in the fixed part of the model and then code it as numeric (continuous) for the random part of the model?
3) Using load as continuous in both the fixed and the random part of the model. In this case, however, I don't know how to make between-group comparisons at specific load values.
Any suggestions are welcome.
Thanks in advance.
Manuel
Hi,
Here the three groups are the Nominal Categories.
Load is increased in a specified fixed value and increasing, hence it is an ordinal variable. Hence it is a fixed factor.
The outcome measures are on a continuous scale, getting repeated measurements under each of the loads.
Keeping these while fitting a mixed model I feel is important.
The factors can appear only once in the model.
You could use a between x within anova with polynomial regression.
https://nmmichalak.github.io/nicholas_michalak/blog_entries/2018/nrg07/nrg07.html
or:
https://stats.stackexchange.com/questions/41654/contrasts-in-repeated-mesures-anova-in-r
For the between x within design with Post Hoc contrasts, see the attached script.
Well did you do the repeats in a continuous manner or just a few? D. Booth
Check https://www.theanalysisfactor.com/models-repeated-measures-continuous-categorical-count-data/
- Badges
- Science topic
- Similar topics
- Psychology