Z values- categorical covariates with multiple levels

[RohitThakur01]

Hi,
Thanks for creating this package.
My question is about how to deal with the categorical covariates in Z.
Some of the covariates that I am dealing with have multiple levels. Should I dummy code this covariate? When I did this, I kept getting this error:

Error in solve.default(t(Z) %*% (Z * d)) :
system is computationally singular: reciprocal condition number = 3.99987e-19

Thanks in advance for the reply.

[pcarbo]

@RohitThakur01 Here's an example of how to create binary covariates from a categorical variable (factor) with three categories (levels):

data(iris)
Z <- model.matrix(~Species - 1,iris)
Z <- Z[,1:2]

Note that if a factor has 3 levels, you only need 2 binary variables, not 3. If you get an error message about a singular matrix, it must be because the rank of cbind(1,Z) is less than 1 + the number of columns of Z. You can check this with qr(cbind(1,Z)). (Here, the 1 is the intercept, which is always included in varbvs.)

Does that help?