I didn’t get a chance to show you the logistic regression for the Kager/Pater data. As discussed in Crawley ch. 14, since these are proportion data, a logistic model is appropriate. Crawley (p. 255) in fact discusses an analogous case with two categorical explanatory variables. Here are the KP data formatted appropriately. Try to construct the analysis yourself based on the Crawley example, or run mine. This model does not include an error term for repeated measures – it looks like we’ll have to follow fashion and use a mixed effects model…

On the subject of stats, I may not have told you that I posted a nice explanation of planned and unplanned contrasts that Wendell sent me a couple of weeks ago – see the page to the left. Also, I’ve just posted a flow chart for choosing a statistical test that Kevin sent – see the stats resources page.