Barbara,
The best way to do this is to model the independent variables at both
the individual and group levels. You would do this by centering the
independents around the mean at the individual level and modeling group
intercepts at the group level. When this is done you model the
individual and group level variance at the level at which they appear
and thus avoid biased estimates. However, not all individual level
variables (or at least measures) display significant variance at the
group level. In other words, if your independent variables don't vary
across groups (group means that is) then you can't estimate the impact
on group level dependent variables. I think this approach is ideal but
so far it's sparsely applied. Personally, I prefer using the mplus
software for this but there are many other options. A less sound but
acceptable (at least so far) approach is to aggregate the individual
level variables to the group if this can be justified given your ICC
values. The following paper may be of interest for you:
David A. Hofmann & Mark B. Gavin: Centering Decisions in Hierarchical
Linear Models: Implications for Research in Organizations (JOM 1998).
Good luck with your research!
Best regards,
--
Mathias Höglund
Management and Organization
mathias.hoglund@hanken.fi
Tel: +358 (0)40 3521 454
Hanken Svenska handelshögskolan
Hanken School of Economics
Arkadiankatu 22
PO Box 479, 00101 HELSINKI, FINLAND
http://www.hanken.fi
Date: Sun, 25 Apr 2010 14:02:14 +0200
From: Arnold Bakker<
bakker@FSW.EUR.NL>
Subject: Re: HLM question
MIME-Version: 1.0
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Dear Barbara,
I was thinking about this reference as well, but must admit that the
content of that paper is complex. Please let me know if you manage to
model this.
Also, note that Schnijders& Bosker (1999; Multilevel Analysis: An
introduction to basic and advanced multilevel modeling, Sage Publishers)
also indicate that the effect of individual on group makes theoretically
and statistically sense.
Best wishes,
Arnold Bakker
> Barbara-
>
> This could be of interest to you:
> Croon, M. A.& van Veldhoven, M. (2007): Predicting group-level
> outcome variables from variables measured at the individual level: A
> latent variable multilevel model. Psychological Methods, 12(1), 45-57.
> (As Corinne noted, aggregation to the higher level is also an option.)
>
> hth,
>
> torsten
>
>
> On 24.04.2010 22:42, Bendersky, Corinne wrote: I'm afraid your DV has
> to be at the lowest level of analysis so you can't test individual
> behaviors on group level outcomes. You have to aggregate the
> individuals. I find it often works best to include both the mean and
> standard deviation. Best of luck, Corinne
> ________________________________________ From: Organizational Behavior
> Division Listserv [
OB@AOMLISTS.PACE.EDU] On Behalf Of Jing
> [
barbarahan_cn@HOTMAIL.COM] Sent: Saturday, April 24, 2010 9:50 AM To:
>
OB@AOMLISTS.PACE.EDU Subject: [OB-LIST] HLM question Dear colleagues,
> Can anybody tells me whether it's possible to test the influence of
> individual behavior on group level or organizational level outcome
> through HLM analysis? If not, any other method to testify? or any
> recommended papers or books where they mentioned this? Many many thanks,
> Barbara
--------------------------------Disclaimer--------------------------------
De informatie verzonden in dit e-mail bericht inclusief de bijlage(n) is
vertrouwelijk en is uitsluitend bestemd voor de geadresseerde van dit
bericht. Lees verder:
http://www.eur.nl/email-disclaimer
The information in this e-mail message is confidential and may be legally
privileged. Read more:
http://www.eur.nl/english/email-disclaimer