s********8
发帖数: 50 | 1 I am thinking all the times about this question when cooking, waiting for
shuttle, taking shower ...., but no hint at all. So ask for help here .
I want to do case-control(1:n) matching on age, medical score and gender.
but frequency mathcing on gender is required.
I just listed what I thought about for criticism .
Method I : if frequency mathcing on gender first, then get two control
pools for female and male respectively.
Method IA: for each pool for gender, I do individual matching on age and
medical score. next I combine two sets of controls from these two pools .
However it is actually individual matching on gender.
Method1B: after I got controls after frequency mathcing on gender at 1:m
(m>>n), I do individual matching on age and medical score at 1:N. but when
I do individual matching, how can I keep the results of frequency matching ?
Method II: I do individual matching on age and medical score first, for
each combination of age*medical score, do frequency matching for gender ? it
is not practical. also it is hard to get 1:N matching for each level of age
*medical score.
any hint ? I do appreciate it.
Happy Valentine's day ! | m****e
发帖数: 255 | 2 My 2 cents.
Regression tree first for Method II. At least you have a smaller number of
partitions of the data. | s********8
发帖数: 50 | 3 more cents from othe audience ? 3X
【在 m****e 的大作中提到】
: My 2 cents.
: Regression tree first for Method II. At least you have a smaller number of
: partitions of the data.
|
|
