Models with only sigma of a response

[Matthew McGee]

I know how to fit a model where the mean and sigma of variable Y are affected by variable X.

bf(Y ~ X, sigma ~ X)

but what if you only want to estimate X's effect on sigma?

Is (a) there a way to specify this and (b) would this reduce computation time at all as compared to estimating both?

[Paul Buerkner]

bf(Y ~ 1, sigma ~ X)

The computation time reduction will likely be rather marginal.

--
You received this message because you are subscribed to the Google Groups "brms-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to brms-users+unsubscribe@googlegroups.com.
To post to this group, send email to brms-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/brms-users/32446f56-6233-42c3-9375-efe2b0894272%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Matthew McGee]

Thanks Paul!

Interestingly, for cases where one is using "cov_ranef", the difference in speed appears to be almost an order of magnitude faster. Will dig into this a bit more to see how general this really is.

On Monday, April 16, 2018 at 7:46:13 PM UTC+10, Paul Buerkner wrote:

bf(Y ~ 1, sigma ~ X)

The computation time reduction will likely be rather marginal.

2018-04-16 11:37 GMT+02:00 Matthew McGee <mcgee....@gmail.com>:

I know how to fit a model where the mean and sigma of variable Y are affected by variable X.

bf(Y ~ X, sigma ~ X)

but what if you only want to estimate X's effect on sigma?

Is (a) there a way to specify this and (b) would this reduce computation time at all as compared to estimating both?

--
You received this message because you are subscribed to the Google Groups "brms-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to brms-users+...@googlegroups.com.

[Paul Buerkner]

If "X" is complex (with random effects etc.) then of course it may make a huge difference. I thought "X" was just an ordinary predictor.

To unsubscribe from this group and stop receiving emails from it, send an email to brms-users+unsubscribe@googlegroups.com.

To post to this group, send email to brms-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/brms-users/e03467c9-4f75-47f2-ba6d-571d9065bd48%40googlegroups.com.