on implementing 2SRI/CF through Prof Terza's code - Matrix multiplication is not happening due to error 3200

[steny rapheal]

Hi
I am working with DHS data on implementing the Control Function approach following Prof Terza's guide. I use STATA 17.

Endogenous regressor: Women's Occupation
Instrument: Cluster Average of Women's Occupation
Outcome Variable: IPV
I am doing probit estimation, where most counterfactuals are categorical variables.
Number of observations: 9342

I followed Terza's code

https://pmc.ncbi.nlm.nih.gov/articles/PMC5980262/

- ran both stages, generated residuals, saved X and W matrices

The following step of matrix multiplication runs into conformability error (3200) in my context:

/*step i*/
mata: gradbeta=exp(X*betahat):*X
mata: gradalpha=-Bu:*normalden(X*betahat):*exp(W*alphahat):*W

I checked for dimensions of the matrices and realised the mismatch:

. display "Dimensions of X: " n_rows_X "x" n_cols_X
Dimensions of X: 9342x23

. display "Dimensions of betahat: " n_rows_betahat "x" n_cols_betahat
Dimensions of betahat: 83x1

I assume the dimensional mismatch between 23 and 83 is the issue. Is this because of the categorical variables involved in the estimation? - When I include categorical variables, the model will estimate one coefficient for each dummy variable. If I have many categorical variables with multiple categories, the total number of estimated coefficients will increase beyond the number of original variables in the matrix X.


How do I work around this and achieve coformability that ensures matrix multiplication? Any help will be much appreciated.

Thank you