[R-meta] stepadj argument ignored in an rma.mv model

[Viechtbauer, Wolfgang (SP)]

Fully agree. I've seen things like:

some_modeling_function(mydata$y ~ mydata$x1 + mydata$x2, data=mydata)

in other places, which is really bad practice. The 'data' argument is there to take care of this for you.

Best,
Wolfgang

-----Original Message-----
From: Michael Dewey [mailto:lists using dewey.myzen.co.uk] 
Sent: Friday, 10 May, 2019 12:35
To: Viechtbauer, Wolfgang (SP); Danka Puric
Cc: r-sig-meta-analysis using r-project.org
Subject: Re: [R-meta] stepadj argument ignored in an rma.mv model

Dear Wolfgang

I seem to remember being bitten by this feature in other functions 
outside metafor so I suspect the answer is really that it is not best 
practice to do that anywhere but to always rely on the data= parameter. 
I think telling people not to do it in the documentation and throwing an 
error if they do is a better use of your time.

Michael

On 10/05/2019 11:20, Viechtbauer, Wolfgang (SP) wrote:
> Just as a follow-up for those who are interested.
> 
> The problem here stems from using a call such as:
> 
> res <- rma.mv(yi, V, random = ~ 1 | mydata$study, data=mydata)
> 
> namely, using 'mydata$' in the 'random' argument. This is actually superfluous (rma.mv() will find 'study' inside 'mydata' with no problems). While the model fit works, this causes some problems with profile(). Making this work is not an easy fix. But the obvious fix is to just fit the model with:
> 
> res <- rma.mv(yi, V, random = ~ 1 | study, data=mydata)
> 
> and then profile(res) works just fine.
> 
> I have now actually added a check inside rma.mv() that catches uses of $ in the random argument and, if so, issues an error. If I come up with a solution for making things work even with $ inside 'random', I'll remove this check.
> 
> Best,
> Wolfgang
> 
> -----Original Message-----
> From: Viechtbauer, Wolfgang (SP)
> Sent: Tuesday, 07 May, 2019 23:34
> To: 'Danka Puric'
> Cc: r-sig-meta-analysis using r-project.org
> Subject: RE: [R-meta] stepadj argument ignored in an rma.mv model
> 
> Well, that fails spectacularly. Not a single successful fit across all 20 sigma^2 values (all ll values are NA). This is a bit surprising, since the actual estimate of sigma^2 is 0 and even for this value the profiling fails. I can't tell if this is a problem with the profile() function or due to a model that is so overparameterized that all bets are off. Would you be willing to send me the data so I can do some more testing to figure out what is going on?
> 
> Best,
> Wolfgang