ProbABEL with 2 interaction terms, or G x (categorical) E

Questions about ProbABEL are welcome here.
Forum rules
Please remember not to post any sensitive data on this public forum.
The first few posts of newly registered users will be moderated in order to filter out any spammers.

When get a solution to the problem you posted, please change the topic name (e.g. from "how to ..." to "[SOLVED] how to ..."). This will make it easier for the community to follow the posts yet to be attended.
aron
Posts: 2
Joined: Fri Jan 28, 2011 10:15 pm

ProbABEL with 2 interaction terms, or G x (categorical) E

Postby aron » Mon Feb 14, 2011 3:50 pm

Hi :
Let me use an example to explain my question here. To model an linear regression analysis "pheno : SNP + SNP x Alcohol + Alcohol + age + sex + error, where Alcohol has 3 levels", I prepare a phenotype file as "id, pheno, alcohol1, alcohol2, age, sex", then using "--interaction =1 --interaction=2" in the analysis. However, as I found out, ProbABEL was not able to handle 2 terms both interacting with SNPs by using 2 separate "--interaction" options.
I am curious if ProbABEL can handle a (non-interaction) covariate variable that is categorical ? Certainly I can create dummy variables for each categorical covariate (as shown above) to handle the task. But perhaps there is a trick/tip somewhere that I am not aware of ? I am asking this since if ProbABEL can model a categorical covariate directly (without the use of surrogate dummy variables), then perhaps I will be able to just add one "--interaction =1" to generate the statistics I wanted?
Your comments are well-appreciated!
Aron

yurii
GenABEL developer
GenABEL developer
Posts: 263
Joined: Fri Jan 21, 2011 5:20 pm

Re: ProbABEL with 2 interaction terms, or G x (categorical)

Postby yurii » Wed Feb 16, 2011 10:30 pm

Aron,

As for categorical covariate, you are right, the only way to get that in ProbABEL is to recode it as a series of binary variables.

Then I thought of (cathegorical E)xSNP, and was going to answer 'no way, sorry', but suddenly realized we have that implemented in our MixABEL.

The only thing you need to do is to re-code your categorical as a series of binaries in rather standard way (e.g. in my example below I code 3 categories of BMI as 'obese'=0/1 and 'overweight''=0/1; the normal-weight being the reference). You need to figure out the most appropriate re-coding yourself.

Here is an example

Code: Select all

library(MixABEL)
data(ge03d2.clean)
overw <- rep(NA,nids(ge03d2.clean))
obese <- rep(NA,nids(ge03d2.clean))
attach(phdata(ge03d2.clean))
# bmi < 25 => obes=overw=0
# 25 < bmi < 30 => overw=1 obese=0
# bmi > 30 => overw=obese=1
overw[which(bmi>=25)] <- 1
overw[which(bmi<25)] <- 0
obese[which(bmi>=30)] <- 1
obese[which(bmi<30)] <- 0
out <- rnorm(nids(ge03d2.clean))
res <- GWFGLS(out ~ sex+age+SNP*obese+SNP*overw,data=ge03d2.clean,ver=2)
res[1:10,]


do not pay attention to lambdas or whatsoever -- this is just a feasibility example, you can do categorical x SNP with MixABEL! Strictly speaking, your outcome is best quantitative :)

Yurii
Note that (Gen)ABELs are dynamically developing; while this post is intended to provide full information at the time of posting, please read on further posts, if any, as the topic may be updated with novel solutions at a later stage.

best regards,
Yurii


Return to “ProbABEL”

Who is online

Users browsing this forum: No registered users and 1 guest