Nested Anova - Use sums instead of mean?

[Daikoro]

Dear all,

At the moment, I'm trying to do a 'Nested Anova'. I have several years (2004 to 2012) and four seasons. So a nested anova is called for correct?

I'm looking at carcass data. My dataset consists of the amount of carcasses found every 10 days. There are a lot of zero values, as I expect higher mortality (at the end) in certain seasons. Anova compares the means of each group, but I think that the 'mean found carcasses' is not really suitable (especially because of the many zero values). I would rather use the 'sum of carcasses' in each season for each year. However, that will result in one value for each subgroup.
As a back-up plan, are there any alternatives for nested anova?

I hope some could shed some light on this matter.
Many thanks in advance!

Kind regards,

Nick

 

[Jake]

It sounds like Poisson regression (LINK) would be appropriate.

If you decide to pursue this, the following textbook may be useful to you: http://www.highstat.com/BGGLM.htm

 

[Daikoro]

Thank you very much for your quick reply! I really appreciate it.

After your suggestion of the Poisson regression, I started to read more about it. The Poisson regression is more used for prediction models, correct?
I didn't mentioned it in my first post, but my goal is to see if there are any differences in mortality between the seasons or between the years. At first, I used an One-way Anova. However, my supervisor suggested a nested anova, because season is nested within year, and a one-way anova was not appropriate.

I am wondering now if the Poisson regression is also suitable for my seasons 'nested' in the years. Or is this not a problem any more?

Sorry for my doubts, but I'm just not certain yet about what to do.

 

[Jake]

The Poisson regression is more used for prediction models, correct?

Not really, no.

I am wondering now if the Poisson regression is also suitable for my seasons 'nested' in the years. Or is this not a problem any more?

Poisson regression can handle datasets with nested structure by including random effects. The textbook I linked to above should discuss this.

 

[Daikoro]

Ah I see. Thank you for your reply. I will check out the textbook in more detail.