Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: metareg with (a) std. betas and (b) sequential or stepwise?

From   Roger Harbord <[email protected]>
To   [email protected], "Garrard, Wendy M." <[email protected]>
Subject   Re: st: metareg with (a) std. betas and (b) sequential or stepwise?
Date   Fri, 14 Jan 2005 11:58:21 -0000

--On 13 January 2005 05:48 -0600 "Garrard, Wendy M." <[email protected]> wrote:

Is it possible to do the following with metareg?
(a) Produce standardized betas with the metareg command?

(b) Enter predictor variables in either sequential or stepwise (a.k.a.
hierarchical) fashion? If so, must it be individual variables, or can
blocks of variables be specified (e.g., design characteristics, then
subject characteristics, etc.)?

My meta-analysis involves mean difference (continuous) effect sizes. I
have tried using the OLS regress options with metareg, but the only
output I produce is error messages. For example:
	metareg Y X1 X2 X3, wsse(seY) bsest(reml) beta
   or	sw metareg Y X1 X2 X3, wsse(seY) bsest(reml)

Can someone advice me if I am using the wrong commands, or suggest
effective work-arounds if these options are not available for metareg?

Thanks very much,

Dear Wendy,

I'm currently finalising and testing an update for the user-written -metareg- package (including a dialog written by Tom Steichen) but i'm afraid neither of the facilities you mention below will be included. This may partly reflect a difference in common practice between medical and social sciences as Stata's -metareg- command was developed by those interested in medical applications. We don't tend to use standardised betas much and stepwise methods are often frowned upon - advice is often to choose a fairly small set of explanatory variables a priori based on subject knowledge.

However it isn't difficult to get standardised betas - simply standardise you predictor variables before entering them into -metareg- .
(see )

-sw- will only work with the estimation commands listed at the top of -help sw-. There's no workaround other doing all the separate regressions by hand and looking at the output at each step.

Best wishes,

Roger Harbord
Department of Social Medicine, University of Bristol

* For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index