cmp - unable to allocate matrix

[Federica Fusi]

Hi,

I am using the cmp command to simultaneously estimate three probit equation such as:

DV₁ = b₀ + b₁ + b₂ + b₃ + b₄ + b₅
DV₂ = b₀ + b₁ + b₂ + b₃ + b₄ + b₅
DV₃ = b₀ + b₁ + b₂ + b₃ + b₄ + b₅

The system converges with no issue and the results are as expected.

However, the system does not converge when I remove some variables. In particular, I am interested in estimating the model including b₁, b₂, and b₃ only in one equation, as shown the model.

DV₁ = b₀ + b₁ + b₄ + b₅
DV₂ = b₀ + b₂ + b₄ + b₅
DV₃ = b₀ + b₃ + b₄ + b₅

The message I receive from stata is the following:

unable to allocate matrix;
You have attempted to create a matrix with too many rows or columns or attempted to fit a model with too many variables.


It seems to be a computational issue rather than a statistical one. Is this correct? I am surprised as I have decreased the number of variables included in the model. Is there a way to address this issue? Other possible causes of this problem that I should be attention to?

Thank you in advance.

[David Roodman]

It is odd that simplifying the model appears to take more memory. Do the equation samples expand a lot when you drop those variables? If you send me exact data and code I can take a look.

[Federica Fusi]

No, the sample remains the same (+/- 5 observations). The dataset doesn't have many missing values.

Thank you for your kind offer. I will privately share the data and code.