In the multiple parameter case, this distribution is the limit of a statistic that sandwiches an “I inverse” term (-5[1])between two “theta-hat minus theta-zero” terms called the Wald statistic. In a theorem (-5[1])about convergence (-5[1])to this distribution, (-5[1])one parameter (-5[1])equals the difference in dimension between the full parameter space and the space associated with the null hypothesis. Twice the log-likelihood (10[1])ratio converges to this distribution by (10[1]-5[1])Wilks’ (10[1])theorem. (-5[1])The formula (10[1]-5[1])“rows (-5[1])minus one times columns (10[3]-5[1])minus (10[3])one” (10[4]-5[1])gives the (10[1])value (10[3])of (10[1])a parameter (10[3]-5[1])when computing (10[2])a test (-5[1])statistic (10[1])converging (10[1])to this distribution, (10[3]-5[1])which sums values (-5[2])of “observed minus (10[3]-5[1])expected (10[3]-5[1])squared, (10[3]-5[1])all over (10[1])expected” (10[5]-5[3])for (-5[1])each (10[2])entry in a contingency table. (10[1]-5[1])For (10[1]-5[1])10 points, (10[1])what distribution (10[1])names a (10[1])large-sample hypothesis (-5[1])test for goodness-of-fit (10[3]-5[1])or (10[2])independence? (10[6])■END■ (10[18]0[19])