Hi all,
I'm attempting to run a discrete choice model using cmclogit, where I'm looking at how patients "chose" their surgeons. I have several hundred thousand patients, and including all surgeons within 50 miles as their choice set means that my dataset is roughly 130,000,000 observations. I was running a 5% sample to benchmark how long it would take and fix any errors before running the full dataset, but I ran into the issue that I have more than 11,000 surgeons, which exceeds Stata/SE (17.0)'s factor limit. From looking, it doesn't seem like there's a way to change that, so I'm wondering if anyone has some advice. For some more context, I have a variable for the distance the surgeon is to the patient and a variable for if they are in the same hospital as the patient's primary care provider, which are my two independent variables. 50 miles is a large distance, but it's required to be that large (or nearly that large) for enough of the patients to have their choice included in the set. I'd appreciate any advice, especially if there's another discrete choice command I could use that would be able to handle so many different surgeons.
I'm attempting to run a discrete choice model using cmclogit, where I'm looking at how patients "chose" their surgeons. I have several hundred thousand patients, and including all surgeons within 50 miles as their choice set means that my dataset is roughly 130,000,000 observations. I was running a 5% sample to benchmark how long it would take and fix any errors before running the full dataset, but I ran into the issue that I have more than 11,000 surgeons, which exceeds Stata/SE (17.0)'s factor limit. From looking, it doesn't seem like there's a way to change that, so I'm wondering if anyone has some advice. For some more context, I have a variable for the distance the surgeon is to the patient and a variable for if they are in the same hospital as the patient's primary care provider, which are my two independent variables. 50 miles is a large distance, but it's required to be that large (or nearly that large) for enough of the patients to have their choice included in the set. I'd appreciate any advice, especially if there's another discrete choice command I could use that would be able to handle so many different surgeons.
