[R-sig-eco] Partially non-independent ANOVA?

[Chris Howden]

Thanks for the explanation  Michael,

And there I was thinking I'd found a silver bullet, should have know if
it's too good to be true it likely isn't. I've been meaning to buy Pinhero
and Bates, and I think this is just the incentive I needed.

Chris Howden
Founding Partner
Tricky Solutions
Tricky Solutions 4 Tricky Problems
Evidence Based Strategic Development, IP Commercialisation and Innovation,
Data Analysis, Modelling and Training
(mobile) 0410 689 945
(fax / office)
chris at trickysolutions.com.au

Disclaimer: The information in this email and any attachments to it are
confidential and may contain legally privileged information. If you are
not the named or intended recipient, please delete this communication and
contact us immediately. Please note you are not authorised to copy, use or
disclose this communication or any attachments without our consent.
Although this email has been checked by anti-virus software, there is a
risk that email messages may be corrupted or infected by viruses or other
interferences. No responsibility is accepted for such interference. Unless
expressly stated, the views of the writer are not those of the company.
Tricky Solutions always does our best to provide accurate forecasts and
analyses based on the data supplied, however it is possible that some
important predictors were not included in the data sent to us. Information
provided by us should not be solely relied upon when making decisions and
clients should use their own judgement.



-----Original Message-----
From: Dunbar, Michael J. [mailto:mdu at ceh.ac.uk]
Sent: Monday, 27 June 2011 7:15 PM
To: Chris Howden; r-sig-ecology at r-project.org
Subject: RE: [R-sig-eco] Partially non-independent ANOVA?

Hi Chris

lme (LMMs only) can handle unequal variances of residuals at the lowest
level in your hierarchy using variance functions: this isn't really what
you describe below. This is described in Pinhero and Bates, but for
beginners, best to look at Zuur "Mixed Effects Models and Extensions in
Ecology with R" which gives a good overview.

lmer can of course do LMMs and GLMMs, but so far as I know it cannot do
unequal residual variances by groups.

Both can do unequal random effect (i.e. the hierarchical levels above the
lowest level residuals) variances by group, but with anything more than
two groups this does get a bit complicated. You put the factor to the left
of the | in the random part as if it were a random slope: for
interpretation it can be easier to reformulate your factor as indicator
variables of 1's and 0's in separate columns.

Regards
Mike


-----Original Message-----
From: Chris Howden [mailto:chris at trickysolutions.com.au]
Sent: 27 June 2011 06:01
To: Dunbar, Michael J.; Cliff Beall; r-sig-ecology at r-project.org
Subject: RE: [R-sig-eco] Partially non-independent ANOVA?

Hi Mike,

I'm rather new to mixed models, but am on a few lists at the moment to try
and get my head around them. Your comments on using GLMM's on data with
unequal variances gave me a very pleasant surprise! And after thinking
about it I realised it was a fairly obvious use of GLMM's.

Can I just confirm your comments though?

If I was testing some factor that I knew each group had unequal variances
could I test it using something like the following?

lmer(response~factor + (1|factor), data=data)?

Chris Howden
Founding Partner
Tricky Solutions
Tricky Solutions 4 Tricky Problems
Evidence Based Strategic Development, IP Commercialisation and Innovation,
Data Analysis, Modelling and Training
(mobile) 0410 689 945
(fax / office)
chris at trickysolutions.com.au

Disclaimer: The information in this email and any attachments to it are
confidential and may contain legally privileged information. If you are
not the named or intended recipient, please delete this communication and
contact us immediately. Please note you are not authorised to copy, use or
disclose this communication or any attachments without our consent.
Although this email has been checked by anti-virus software, there is a
risk that email messages may be corrupted or infected by viruses or other
interferences. No responsibility is accepted for such interference. Unless
expressly stated, the views of the writer are not those of the company.
Tricky Solutions always does our best to provide accurate forecasts and
analyses based on the data supplied, however it is possible that some
important predictors were not included in the data sent to us. Information
provided by us should not be solely relied upon when making decisions and
clients should use their own judgement.


-----Original Message-----
From: r-sig-ecology-bounces at r-project.org
[mailto:r-sig-ecology-bounces at r-project.org] On Behalf Of Dunbar, Michael
J.
Sent: Thursday, 23 June 2011 10:59 PM
To: Cliff Beall; r-sig-ecology at r-project.org
Subject: Re: [R-sig-eco] Partially non-independent ANOVA?

Dear Cliff

Using a mixed-effects model (also variously termed multilevel or
hierarchical model) could well be what you are looking for. These are
implemented in several R packages.

In your study, in mixed model parlance, patient identifier would be a
random effect (you don't say how many patients), with some patients having
one measurement and some two, and treatment (control or periodontitis)
would be a fixed effect. How or whether you code the sites (which would be
another fixed effect) depends on whether they are of interest.

These approaches are very powerful, but I wouldn't advise just dipping
into them - it can be hard to get things set up correctly if you just want
"a test", they repay careful study. On the bonus side, once you're
familiar with them you start seeing the everything in a hierarchical
mindset!

You can model unequal variances by group, although you need to be careful
in model specification.

Regards
Mike

-----Original Message-----
From: r-sig-ecology-bounces at r-project.org
[mailto:r-sig-ecology-bounces at r-project.org] On Behalf Of Cliff Beall
Sent: 22 June 2011 17:18
To: r-sig-ecology at r-project.org
Subject: [R-sig-eco] Partially non-independent ANOVA?

Dear group-

I was wondering if someone could point me towards the correct statistical
test and R-function for my situation.

I am looking at oral bacteria and have three groups of samples, one from
control subjects, and two that are derived from different sites in
patients with periodontitis. So I have one totally independent group, and
two that have samples that are paired.

I would like to do a test just to say that there are differences between
the three groups, something like an ANOVA, but that deals with the lack of
independence of the two groups.

By the way, for what I am measuring it looks like normality is a
reasonable assumption but equal variances are probably not.

Thanks,

Cliff Beall
Research Assistant Professor
Division of Oral Biology
College of Dentistry
The Ohio State University
beall.3 at osu.edu

_______________________________________________
R-sig-ecology mailing list
R-sig-ecology at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
--
This message (and any attachments) is for the recipient ...{{dropped:10}}