[#6784] maintain non-structural zeros in Khatri-Rao product

Date:
2022-10-23 23:25 Priority:
3

State:
Open

Submitted by:
Ben Bolker (bbolker) Assigned to:
Nobody (None)

Hardware:
None Product:
None

Operating System:
None Component:
None

Version:
None Severity:
enhancement

Resolution:
None

URL:

Summary:
maintain non-structural zeros in Khatri-Rao product

Detailed description
Copying from https://github.com/lme4/lme4/issues/671, where there is more discussion ``` library(Matrix) dat <- data.frame(y = 1:3,id = rep(1:2,each=3),t = c(0,1,0,2,3,4)) sm <- fac2sparse(factor(dat$id), to = "d") mm <- model.matrix(~t, data = dat) sm <- KhatriRao(sm, t(mm)) Matrix::t(sm) ``` Results: ``` 6 x 4 sparse Matrix of class "dgCMatrix" 1 1 2 2 1 1 . . . 2 1 1 . . 3 1 . . . 4 . . 1 2 5 . . 1 3 6 . . 1 4 ``` Desired results (i.e., keeping zero values in original `mm` as non-structural): ``` 6 x 4 sparse Matrix of class "dgCMatrix" 1 1 2 2 1 1 0 . . 2 1 1 . . 3 1 0 . . 4 . . 1 2 5 . . 1 3 6 . . 1 4 ```

Comments:

Message
Date: 2022-11-02 18:34 Sender: Ben Bolker I didn't know that -- thanks for pointing it out. I guess I will add `sparseY=TRUE` to the lme4 code calling this (with further back-compatibility hooks ...)
Date: 2022-11-02 18:27 Sender: Mikael Jagan I'm a bit confused. Didn't Martin add an argument 'sparseY' in Matrix 1.5-0, with default TRUE for backwards compatibility? > KhatriRao(sm, t(mm), sparseY = FALSE) 4 x 6 sparse Matrix of class "dgCMatrix" [1,] 1 1 1 . . . [2,] 0 1 0 . . . [3,] . . . 1 1 1 [4,] . . . 2 3 4 But perhaps we also need a 'sparseX' argument? We must deal with kronecker() simultaneously, i.e., kronecker() methods involving denseMatrix or traditional matrices should also get an argument, say 'drop0', specifying how zeros in the dense factor(s) are treated.

No Changes Have Been Made to This Item