Best method of analysis?

[rwild]

I have 900 subjects undergoing up to 5 cycle attempts at getting pregnant

of interest is best method to define if characteristics of the insemination (difficult , **bleep** characteristics, total motile count, etc,catheter type, bleeding, **bleep** characteristics before washing and insemination each cycle

are related to achieving a live birth (censored)

I am trying to decide if GEE is the best approach and if so what specifications for this binary censored outcome ?

Is survival analysis better?

one suggestion is to use a weighted poisson specification in genmod (GEE with poisson and weighted for number of inseminations however this is number of insemination only and doesn't take into consideration reduced fertility over number of cycles attempted

this seems to be a population approach rather than an individual probability of success

complicated because not every patient will continue

some will drop out (frustrated)

some will get pregnant

so the group with first attempt is more fertile than those with 2 tries or 3 tries or 4 tries or 5 tries etc

and of interest is factors influencing success with the technique adjusting for known fertility factors like age , months trying to conceive etc

thoughts?

we have SAS 9.4 for our University of Oklahoma Health Sciences Center

thank you for guidance

rwild

[SASKiwi]

Your question seems more related to choice of statistical method rather than anything directly SAS-related. I'd suggest posting your question on specialist statistics forums especially any biological ones.

You could also raise a SAS Tech Support problem and ask to get in touch with a SAS statistical specialist.

[rwild]

Thank you



How do I get in touch with a SAS statistical specialist?



Thank you



rw

[SASKiwi]

@rwild - submit a Technical Support problem here: https://support.sas.com/ctx/supportform/createForm

[rwild]

How do I contact senior statistician ?

[SASKiwi]

@rwild  - As I explained above, raise the Tech Support problem online and ask for it to be forwarded to a statistical specialist.

You should also consult statistical specialists in your own organisation.

[pau13rown]

sas technical support don't respond to technical statistical Q's like this (they have refused me before for this reason, they condiser it consultancy, fair enough).

i'm not sure why you would consider GEE here. Survival analysis would make more sense (proc phreg) or logistic regression (proc logistic or proc genmod). Need a guesstimate of censoring rate etc to inform the best approach.