Scale and normalize genotypes for GRM.

The function creates a scaling function to be used for linear algebra operations on FBMs through the fun.scaling argument in downstream functions from bigstatsr.

big_scale_grm(center = TRUE, scale = TRUE, M)

Arguments

center	Center?
scale	Scale?
M	A normalization value. It is typically equal to the number of columns in a matrix (for which the scaling function is created).

Value

A function that returns a data.frame with two columns "center" and "scale".

Details

The scaling function works such that the input matrix of raw genotypes is represented as a matrix Z, which cross product is the genetic relationship matrix (GRM), defined as scale(G) scale(G)' / M. In other words, GRM = ZZ'.

Examples

G <- attach_example200()
dim(G)
#> [1] 1500  200

M <- dim(G)[2] # number of columns
fun_sc <- big_scale_grm(M = M)
stats <- do.call(fun_sc, list(G))
str(stats)
#> 'data.frame':	200 obs. of  2 variables:
#>  $ center: num  0.966 1.002 1.005 0.997 0.997 ...
#>  $ scale : num  9.85 10.02 9.84 10 9.92 ...

# compare to the standard scaling function from bigstatsr
fun_sc0 <- big_scale()
stats0 <- do.call(fun_sc0, list(G))
str(stats0)
#> 'data.frame':	200 obs. of  2 variables:
#>  $ center: num  0.966 1.002 1.005 0.997 0.997 ...
#>  $ scale : num  0.697 0.709 0.696 0.707 0.701 ...