search

mcaceresb / stata-pretrends

Power calculations and visualization of pre-trends tests following Roth (2022). (Stata version of the R package of the same name.)
22 stars 12 forks source link

WARNING: Unable to load mvnormalcv() plugin; falling back on mata. #8

Open asun146 opened 1 month ago

asun146 commented 1 month ago

I keep seeing the warning message below whenever I run any command from this package. This doesn't prevent me from obtaining the exact same results as tabulated on the ReadME page, but would there be a way to address it? Thanks!

WARNING: Unable to load mvnormalcv() plugin; falling back on mata. Execution may be excessively slow.

image

jonathandroth commented 1 month ago

The results should be fine, they just may be slow if you have a large number of periods. If it's fast enough, you can just ignore it. If not, Mauricio may be able to help you with package installation

On Sat, Oct 26, 2024 at 11:15 PM Aiqi Sun @.***> wrote:

I keep seeing the warning message below whenever I run any command from this package. This doesn't prevent me from obtaining the exact same results as tabulated on the ReadME page, but would there be a way to address it? Thanks!

WARNING: Unable to load mvnormalcv() plugin; falling back on mata. Execution may be excessively slow.

image.png (view on web) https://github.com/user-attachments/assets/b7965034-f891-49e8-8532-4b01aa857637

— Reply to this email directly, view it on GitHub https://github.com/mcaceresb/stata-pretrends/issues/8, or unsubscribe https://github.com/notifications/unsubscribe-auth/AE6EXFCODPP2R5FEOFX4LS3Z5RLF7AVCNFSM6AAAAABQVMMM7SVHI2DSMVQWIX3LMV43ASLTON2WKOZSGYYTMMRXG4YTOMI . You are receiving this because you are subscribed to this thread.Message ID: @.***>

mcaceresb commented 1 month ago

@asun146 As we noted here, Stata's implementation of the multivariate normal distribution via mvnormalcv() can be very slow. pretrends implements its own version internally in C; the warning message means the C plugin failed to load so pretrends is falling back on mvnormalcv(), which can be slow. As @jonathandroth mentioned, the results should be identical (up to numerical precision), just computationally slow.

PS: Do you happen to be on Windows 11? I've had some bug reports of plugins failing to load on some versions of Windows 11; I can't debug that at the moment but just trying to see if that's the main common thread lately.