Hi,
Can anyone tell me how wrong can your answer be if you model some cross
sectional time-series binary data which is better modelled by some
probability mechanism based on previous observations, eg logit p_ij =
alpha_i0 + alpha_i1*y_i(j-1) + alpha_i2*y_i(j-2), with a simple random
intercept model ie assuming the observations within an individual are
independent?
Thanks in advance,
Tim
|