

Relative vs. absolute differences 

Message/Author 

LSRM posted on Friday, June 07, 2013  2:27 pm



I have a twopart growth curve analysis predicting functional limitations. If the slope is similar for males and females, does that mean the relative difference in the outcome stays the same, or the absolute difference? Is it possible for me to separate out the absolute and relative differences in the outcomes? Is it correct to say: Binary intercept: Females have 2.97 times (e 1.09=2.97) greater odds of having any limitations at baseline compared to males. Continuous intercept: Females have 18.6% (100*0.186) more limitations at baseline among those with limitations. Could I also say that females have 1.20 (e .186 = 1.20) more limitations at baseline. Thank you. 


The statement for the binary intercept (intercept for the binary part of the twopart growth model) seems correct, but for the continuous intercept you don't want to talk in terms of e.186 because the DV is continuous, not binary. Instead you do the usual linear regression interpretation: females are 0.186 higher than males  and relate that to the proper variance (which can be not only the intercept variance but also the residual variance for the outcome at baseline. I don't know what relative and absolute differences are  sounds like a general question suitable to SEMNET. 

Back to top 

