Unbalanced ANOVA with missing treatment/factor combinations

[jwmorley]

Hello, I am analyzing results from an experiment with a single response variable, which is attack speed of a fish predator. The two independent variables are temperature (three levels) and prey size (two levels). Each treatment/factor has the same number of replicates, but I have no data for one of the combinations (zero replicates from the middle temperature/larger prey).

First, I just dropped the corresponding treatment/factor (middle temperature/small prey) and did a fully crossed 2-factor ANOVA. In other words, I just used two temperature treatments. However, I am curious if there is a way to analyze this without dropping that treatment?
I appreciate the help,
Jim

 

[Miner]

Do you have actual temperatures and prey sizes that you are classifying into treatment levels? If so, you should be able to analyze using regression. If not, you need to look at analyzing as a Balanced Incomplete Block Design. Montgomery and Milliken both cover this topic. I recommend Milliken.