Hi Molly
Model 2
alpha_i + beta_k + gamma_j + sigma_ij would have been an acceptable alternative to alpha_ij + beta_k
The alpha_ij term is made up of the combination of alpha_i + gamma_j + sigma_ij
There is a question and answer in Section 3.3, p25 that runs through this with some numbers, which may help.
Model 3
The alpha_ijk term is made up of alpha_i + gamma_j + beta_k+ sigma_ij + delta_jk + theta_ik + lambda_ijk
All the values have been combined into alpha_ijk, this includes all the effects
Number of parameters
1 factor alone parameters = number of categories eg YO = 2
as you add factors + (n-1) where n is the number of categories eg parameters in YO+FS = 2 + (2-1)
as you add interactions +(n-1)(m-1) where n is the number of categories for first factor and m is the number of categories for second factor eg in YO+FS+YO.FS = 2 + (2-1) + (2-1)(2-1) = 4
To solve the parameters we have to combine some of the constants, so that is why they are less than you might have expected, again the question and answer in Chapter 13, Section 3.3 on p25 may help you see this in practice.
Thanks
Andrea
Last edited: Mar 14, 2023