looking for a relationship between multiple catagorical data sets

[jerryb]

I have recently come across a problem that is clearly out of my level of background, so perhap someone can help:

I wish to determine if condition 1 is related to conditions 2,3,4 etc. its not clear to me just yet how many conditions will ultimately be involved. I would like to get a sense of how to treat this. for each condition the is either a true or false outcome.

let me give an example that may help: I wish to know if the proportion of instances of condition 1 being present is greater or lesser when condition 2 is also present, then if it is greater when cond 3 is present and so on. now that part i can handle. but i would also like to know if the proportion of cond 1 is greater when both conditions 2 and 3 are true, or both cond 3 and 4, or both 2 and 4, or even if three conditions are present (ie cond's 2, 3 and 5).

I would think this has been done often, i just need a push in the right direction.

cheers
jerry

[vinux]

Hi Jerry,
Could you just explain more on conditions? and I guess it is a multinomial variable.

Richie

[jerryb]

Hey Richie,

let me try an example:

suppose that for a subject we know that condition a is true, b is false, c is true, d is true, (all the conditions are binomial, eg true or false)

now we collect the same data on many subjects so that we find the proportion with a is .2, with b is .1 with c is .4 with d is .1; now i would like to be able to establish if a subject is more likely to have condition a be true if they have condition c true? or b true, or d true, etc.

another way of thinking?? maybe we want to test Ho: the proportion with "a" and "b" is the same as the proportion with "a" for all combinations of "a" and one, or two or more conditions.

thanks
jerry

[jerryb]

this is not a multinomial variable, BTW

I am really trying to test wether there is a relationship between multiple binomial variables.

thanks again
jerry

[vinux]

I guess pairwise we can establish the relationship.
Suppose there is only a & b. In that case we can summarize the data into 2*2 table ( a=true b= true , a=false b = true , ... )
Then we can test both McNemar's paired test & chi-square test( independence).

[jerryb]

I found a fellow to assist with this, I'll let you know what we come up with.
jerry