YouTube continuous by continuous

[Dave Airey]

This StataCorp video needs updating. It shows a neat profiler plot, but states at the end that the pattern is indicative of an interaction. The model clearly did not have an interaction earlier in the video. Or I need a new prescription for my glasses.

https://www.youtube.com/watch?v=QFROtui_OyM

[Scott Merryman]

From 2:08 in the video:

[Dave Airey]

I meant the included interaction term was not significant.

[Nathan E. Fosse]

Interestingly, I had written some code using those data (probably in reference to the video). It seems to be the case!

use "http://www.stata-press.com/data/r12/nhanes2.dta" , clear
reg bpsystol c.age##c.bmi
sum age if e(sample) == 1
sum bmi if e(sample) == 1
loc min = round(r(min),1)
di `min'
loc max = round(r(max),1)
di `max'

margins, dydx(age) at(bmi=(`min'(1)`max')) vsquish

marginsplot

[Chuck Huber (StataCorp)]

Yes, that was an unfortunate oversight on my part. I was more focused on demonstrating the mechanics and I liked the way the final graph looked. I should have chosen an example where the interaction term was statistically significant. I apologize for any confusion and I'll re-record the video at some point.

[Dave Airey]

I enjoy your videos!

[Chuck Huber (StataCorp)]

Thanks - glad you find them useful!

[Nick Cox]

Chuck was overlooked again in the Oscars. Shame.