Question

multiple interaction factors

0

Entering edit mode

bobyboby • 0

@bobyboby-7062

Last seen 11.1 years ago

Germany

Dear limma users,

I would like to know if it possible to include two random interacting factors in the design matrix(like block:tissue+tissue:genotype) and use in the lmFit() function.

thanks for the help.

regards,

Boby

limma • 774 views

ADD COMMENT • link updated 11.2 years ago by James W. MacDonald 68k • written 11.2 years ago by bobyboby • 0

score 2 · Answer 1 · 2014-11-18

2

Entering edit mode

James W. MacDonald 68k

@james-w-macdonald-5106

Last seen 1 hour ago

United States

Any linear model that you would be able to specify for lm() will also work for lmFit(). In other words, as long as your design is valid, you can pass it to lmFit() and get the expected results.

ADD COMMENT • link 11.2 years ago James W. MacDonald 68k

0

Entering edit mode

As a side note, if your design matrix is not of full column rank, lmFit will yield a warning telling you that some coefficients are not estimable. This doesn't change lmFit's behaviour for the estimable coefficients, but it may be problematic, e.g., if it affects the parametrization of your DE contrasts.

ADD REPLY • link 11.1 years ago Aaron Lun ★ 29k