If we want more covariates, do we add rows or more code?

ISG Workshop 2026 Day 3
1

If we want more covariates, do we add rows or more code?

Or I mean more lines of code?

2

you might want to add more lines in the script that create additional objects

rather than extra rows in a matrix

3

So if I wanted to add one for country I would set up all four lines of that code for country?

4

you can follow the came pattern so

mean = intercept + b*sex + b*age + b*season etc

5

With multiple covariates, you can create a regression equation like Sarah showed.

You can also create a matrix expression like

mean = B %*% x

where mean is a vector of expected means, B is a matrix of regression coefficients, and x is a vector of definition variables.

1 Like
6

Could you give some example code, I’m having trouble visualize that

7

I will warn you, covarying out country can get gnarly…if you have k countries represented in your dataset, you will need k-1 dummy variables (or however you want to code it).

8

I was just giving an example, I just want to know how to set up several covariates. As an alternative to regressing them out first

9

Here’s some OpenMx-ish code that does the matrix approach

mxMatrix('Full', nrow=3, ncol=1, values=c(1, NA, NA), labels=c(NA, 'data.Covariate1', 'data.Covariate2'), name='x')

mxMatrix('Full', nrow=2, ncol=3, free=TRUE, name='B')

mxAlgebra(B %*% x, name='expectedMean') # 2 rows, 1 column