module Model where import Probability import System.Environment import Data.Frame -- Ideally, the categories and their weights would be exchangeable! -- Currently if the first category is bad, there is not a good way -- to eliminate it. cluster_dist = do mean <- prior $ normal 0 10 prec <- prior $ gamma 2 1 let sigma = 1/prec return (mean,sigma) model xs = do let n_points = length xs alpha <- prior $ gamma 0.5 10.0 params <- prior $ dp n_points alpha cluster_dist observe xs $ independent [normal mean sigma | (mean,sigma) <- params] let loggers = ["alpha" %=% alpha, "params" %=% params] return loggers main logDir = do [filename] <- getArgs xtable <- readTable filename let xs = xtable $$ "x" :: [Double] return $ model xs