Question

1. Hoem (1987, International Statistical Review 55) considered the association between marital status and mortality of young German men using the below dataset. Table 1: Number of deaths and person years of German males by age and marital status Deaths Person Years Single Married Total Single Married Total 457 91,444 8,556100,000 448 86,83512,70899,543 43975,89223,203 99,095 433 63,24135,41598,656 425 52.023 46.207 98,230 413 42,12355,67597,798 400 36,91560,47097,385 392 32,215 64,770 96,985 Total 2,4859223,407 480,688 307,004 787,692 433 412 337 331 287 27242 215 192 24 36 23 24 25 26 102 138 28 29 185 200 (a) Specify a model for the mortality rate, including an intercept term, age group effects, and marital status effect. Assume that the marital status effect is propor- tional over age. (b) Write the likelihood function for the parameters in the model you specified in (a) (c) Show how the regression coefficient parameter for marital status can be interpreted n terms of a rate ratio. (d) Fit the model and interpret the results. (e) Calculate the expected number of events in each age/marital status category Compare the expected numbers to observed event counts to assess the overall model fit. What statistical test can you use for this? (f) Fit also a model that allows for interaction between marital status and age. What can you say about the model fit now?

Answer 1