Dear group:
I am running logistic regression. I have one dichtomous
predictor
(job strain) I am looking at and adjusting for several confouders,
some continuous some categorical (age, gender, hours, etc). .
In this data, I also have a group of 5 specific work tasks that I
would like to adjust for as well so that job strain is independant of
the numbers of specific tasks that subjects performed. These tasks
were sampled as ordinal frequencies (0/day, 1-5/day, 6-10/day 11-15/
day, 16+ per day). They can not be entered in the same model
because
subsets of them are highly related. 3 of them load highly on one
factor and 2 of them load highly on another (none overlap).
Conceptually the factors make sense and none of them measure the same
thing in terms of the tasks invovled. So.....should I just total the
related items and make 2 scales and enter them, or should i use the
factor scores in the regression analysis to adjust for the two
dimensions. I am not particularly interested in interpreting the
coefficients from either the factor scores or the totaled scales.
Thanks!!!!
|
