Date:         Sat, 16 Jan 2010 19:07:10 -0800
Reply-To:     Jeremy Miles <jeremy.miles@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Jeremy Miles <jeremy.miles@GMAIL.COM>
Subject:      Re: multilevel modeling
Comments: To: Ching <kcwong5@gmail.com>
In-Reply-To:  <b06cc876-d74c-49d0-a200-7d84095918dd@m3g2000yqf.googlegroups.com>
Content-Type: text/plain; charset=ISO-8859-1

If the data have three levels (e.g. children in classes in schools)
you need to incorporate those three levels to get correct parameter
estimates.   You can't just ignore one.  However, you can collapse to
the (say) class level if you are not interested in individual level
covariate effects (I'm not sure what you mean by determinants -
sorry).

Jeremy





2010/1/16 Ching <kcwong5@gmail.com>:
> Hi Eli and everyone.
>
> Sorry about that.
> I'm wondering whether I should model my response variable using 2-
> level or 3-level multilevel model given that my interest is to find
> out the determinants.
>
> Many thanks in advance. Have a great day.
>



--
Jeremy Miles
Psychology Research Methods Wiki: www.researchmethodsinpsychology.com