|
What are the advantages of log linear representation vs table representation of a factor? A factor includes a set of random variables X1 to Xn, and state the probability that each set realization of X1... Xn as (x1, .. xn) happens. Tabel representation: list all possibilities of variables X1 to Xn and put a probability value there. Log linear model: represent using indicator function, only give out to important case. What I am thinking is that log linear model is the same as table representation, except that it skips unimportant realization. Is that true? |
