Re: ANOVA, GLM and missing values

Andy Spragg wrote: 
In the software implementations of ANOVA that I have seen, missing
values are not allowed. To analyse ANOVA-type data with missing values
I have to use GLM, which ANOVA reduces to under certain circumstances.


Having now read a little about GLM, I am unable to see why it should
allow missing values when ANOVA apparently does not. As far as I can
see, the generalization from ANOVA lies in the conceot of a link
function, and allowing this to be something other than the identity
function ushers in a new dimension of modelling possibilities.
However, when the link function is the identity function, GLM is
supposed to reduce back to ANOVA ...

So I fail to see what else GLM brings to the party which allows
missing values to be tolerated. I'd be grateful for any insights.


Most standard implementations of repeated measures ANOVA require one case (i.e., row of data) per subject, and the entire case is dropped if any of the repeated measures are missing. Is *that* what you're talking about? In cases like that, you can restructure the data to have one row per observation (so multiple rows per subject), include subject explcitily as a factor, and cajole your software into doing a repeated measures ANOVA using the between-Ss procedure (or GLM).

I would not say that missing values are "tolerated", because they are still dropped. The difference is that only the observations that are missing are dropped, not the entire subject. In SPSS, the loss of balance can result in some changes to the error terms (i.e., the MS may be the linear combination of the usual MS plus one or more others).

--
Bruce Weaver
bweaver@xxxxxxxxxxxx
www.angelfire.com/wv/bwhomedir
.

Relevant Pages

  • ANOVA, GLM and missing values
    ... In the software implementations of ANOVA that I have seen, missing ... I have to use GLM, which ANOVA reduces to under certain circumstances. ... However, when the link function is the identity function, GLM is ...
    (sci.stat.math)
  • ANOVA with missing values
    ... Can I run a single or multiple factor ANOVA with replication in Excel when i am missing one or more data points? ... Please help - I cannot figure out how to run an ANOVA when I do not have equally sized groups, or when I have to exclude a sample from one of the groups. ...
    (microsoft.public.excel.misc)
  • ANOVA effect size
    ... I'm currently running a 3-way ANOVA and my sample size is very large ... accounted for X% of our explainable variance". ... Is it as simple as adding up the MStreatments associated with ... like that should be a meaningful measure, but perhaps I'm missing ...
    (sci.stat.consult)
  • Re: comparing slopes for significant difference
    ... One of the individuals is missing a lot of data at the ... > like to constrain the regression to pass through the origin. ... but there is no option to do this with a GLM with factors. ... > the slope implies that the rate of change is not constant, ...
    (sci.stat.edu)
  • Re: comparing slopes for significant difference
    ... One of the individuals is missing a lot of data at the ... > like to constrain the regression to pass through the origin. ... but there is no option to do this with a GLM with factors. ... > the slope implies that the rate of change is not constant, ...
    (sci.stat.math)