Dear all,

 I'm running a multi-level model to estimate the average support rate for a movement in each of the seven area. By using the beta model, I got the parameters, such as alpha, beta, and theta as the support rate. My colleague suggests that I should divid the dataset into half so I can improve the estimates of the support rate  in the second half according to the estimates driven from the first half. In addition to put estimated alpha and beta, what can I do with the theta? Your responses will be greatly appriciated.

Chia-hung Tsai
Election Study Center
National Chengchi University
Taipei, Taiwan