What is the correct answer here?

Question 9 For a GLM consider comparing a submodel {model A} to a full model {model B}. - Hg is "model A is appropriate" whereas. H1 is "the full model ts the data better". - The test statistic is the difference in residual deviance between the two models - we will call this the deviance change. - The p-value is usually calculated using a chi-squared with degrees of freedom equal to the number of additional parameters in the full model as the sampling distribution for the deviance change. In some situations we may wish to use simulations to create an empirical sampling distribuon for the deviance change. Consider 3 possible ways of doing this: A. PBS-1: Use a parametric bootstrap where new data is simulating using tted model A. For each simulated dataset, t both models and record the deviance change. B. PBS-2: Use a parametric bootstrap where new data is simulating using tted model B. For each simulated dataset. t both models and record the deviance change. C. NPBS: Use a nonparametric bootstrap where data is simulated by sampling {with replacement} from the original data. For each bootstrap sample. t both models and record the deviance change. Which of these approaches produces a valid sampling distribution for the deviance change? \f