Interaction Test between SNP * Covariate for a Binary Outcome

Feature Requests
1

Hello,

I found that hail can do interaction test for a continuous outcome using the function, hl.agg.linreg.
Can we do the interaction test between snp and covariate for a binary outcome?

Thanks,
Jaeyoon

2

Yes, this is possible Jaeyoon. You’d put your multiplicative interaction term into your covariates: https://hail.is/docs/0.2/methods/stats.html#hail.methods.logistic_regression_rows

3

Ack, unfortunately this is not so simple. The covariates expected in hl.logistic_regression_rows must be column fields, they cannot be entry fields. We are building infrastructure that will make it possible to write logistic regression models like this on Hail’s python interface, but it won’t be ready for some months.

4

That’s good to know! Some months from now :slight_smile:

5

Thank you very much for answering my question. I appreciate it.

Best,

Jaeyoon

6

Hi, do you have any updates on running an interaction test for binary outcomes (logistic regression).

7

@tpoterba Has there been any progress on a solution for an interaction test?

8

We have all the pieces to do this now. We have an implementation of logistic regression that is entirely written at the Python-level using the Hail expression language. Someone needs to get in there and sort out adding more covariates to the X matrix used in the tests. It’s not something I have cycles for in the short term.