Analysis of a matched case-control study

#1
Please, i need help on my analysis. It is a 1-1 matched case-control studies. I am willing to pay for the guidance. My deadline is friday.
 
#2
Hi Madonna!! Outline your study a little more - so it's a quantitative study, using an existing database? What is the topic - it is in the hard sciences or social sciences? Do you have two groups to compare? What are your independent variables (the ones you can manipulate in the data like male/female, years, age, etc) and the dependent variables (the ones that are acted upon). These answers will help - are you trying to figure out which tests to run? here are two good sites with calculators - http://www.ats.ucla.edu/stat/sas/whatstat/ and http://www.microsiris.com/Statistica...ee/default.htm
 
#3
Dear Madonna, hi and welcome to TalkStats.

You don't need to pay. Instead, please do two completely FREE things:
1. Please click the Thanks button below every post you receive (on the lower left cornet of each comment of others).
2. Please come back and contribute to the community by kindly helping the others who know less than you.

The above two items completely suffice :)

By the way, purplelily asked very good question (quoted below). Answering them will clarify many things

Hi Madonna!! Outline your study a little more - so it's a quantitative study, using an existing database? What is the topic - it is in the hard sciences or social sciences? Do you have two groups to compare? What are your independent variables (the ones you can manipulate in the data like male/female, years, age, etc) and the dependent variables (the ones that are acted upon). These answers will help - are you trying to figure out which tests to run? here are two good sites with calculators - http://www.ats.ucla.edu/stat/sas/whatstat/ and http://www.microsiris.com/Statistica...ee/default.htm
You should also tell us about your sample size, how did you match your subjects? and any other detail of your thesis.
 
#4
My sample size is 320. 160 cases and 160 controls. They were matched individually by age. The study is on risk factors associated with secondary infertility. My dependent variable is the group status; case or control, entered as 1or 2. My independent variables include abortion, STI, BMI, miscarriage, genetic e.t.c
 

spunky

Smelly poop man with doo doo pants.
#6
You need to run a conditional binary logistic regression. (or other more sophisticated approaches like multilevel ones).
and i'd add that probably some linear contrasts between the control/case groups would have to be done in order to see where the differences are in terms of risk factors. yeah, a logistic regression with a properly-specified matrix of contrasts sounds right...
 
#7
Thanks Victor, cases already presented at the clinic wiyh secondary infertility. The controls were pregnant women of same age. I know about conditional logistic regression but someone said i have to do mcnemar's chi-square anx odds ratio. My question is in what order? Is it mcnemar's chisquare - odds ratio- regression? I want to know.
Thank you.
 
#9
Thanks for clarifying Madonna :)

Could you also tell us how did you measure the secondary infertility? is it a matter of Yes or No? (I guess so)

So you have 160 infertile people and 160 fertile people. You want to see which risk factors might contribute to infertility?

There are sophisticated and simple approaches for this. The sophisticated ones are better, but I don't think you can learn and do them within a very short time unless you have some background. The simple version is to compare your risk factors between the two groups. For this purpose, you should tell us more details on your variables:
Abortion: How did you record this? Yes or No? or the number of abortions ever done?
miscarriage: the same as above.
Genetics: what is the type of this variable? Some "yes or no"? or what else?
BMI seems to be continuous.
STI?
etc.?

Please clarify the above items very carefully, so that we can decide on the proper tests.

About your second question, you should refresh your page every now and then. Sometimes we answer very fast, and sometimes a little patience might be needed.

So then do a McNemar between the two groups (for each independent variable, do a separate McNemar). It is very different from regression... If you can do a conditional logistic regression, then do that too. the latter is much better than McNemar, but McNemar is as well valid.