R square read with or without confounder in the analysis

#1
Hey,

My bachelor thesis is about fatigue, nausea and cognitive decline in patients who have head and neck radiation. For example, the dependent variable is fatigue, looked at the difference score (score of fatigue 6 months after radiation (T1) - the start of radiation (T0) score.

The multivariable analysis included:
* T0, as a separate variable, due to the expectation that the relationship between the variable and the endpoint Δfatigue may be different between patients with a low or high score at T0
* Chemotherapy
* WHO-score, as a confounder

My questions are:
- Should I look at the adjusted R2 when all variables are included in the analysis, or look at the adjusted R2 without the confounder in the analysis?

And if you want to look at the R2 of all variables in the analysis (including the confounder), the adjusted R2 is 26%
- Does this indicate that the difference between the patients, whether or not fatigue (the declared variance), is explained by 26% by chemotherapy alone? Because the WHO-score in the analysis is taken as a confounder?


Thanks for the answers.

Regards,
Lisa
 

hlsmith

Omega Contributor
#2
You should run your full model, not sure why you want to run it without the confounder?

Most programs allow you to request the partial r-sq contributions for variable (eta-square). Have you established WHO as a confounder,does a variable's effect change when it is included and excluded?
 
#3
You should run your full model, not sure why you want to run it without the confounder?

Most programs allow you to request the partial r-sq contributions for variable (eta-square). Have you established WHO as a confounder,does a variable's effect change when it is included and excluded?
Thank you for the quick answer!
My stats skills are not so good. But i thought that, if the confounder was in the analysis that the difference between the patients, whether or not fatigue, is explained 26% by chemotherapy and thus also the WHO?

What do you mean with partial r-square?

The regressioncoefficient of chemotherapie differs by more than 10%, if i add de WHO-score.