The data for the low-birth-weight study are described in Section 1.6.2. These data are used in Section 8.2 to illustrate ordinal logistic regression models via the four-category outcome BWT4,
However, in this problem use the outcome variable BWT4 and fit the multinomial or baseline logistic regression model with BWT4 = 0 as the referent outcome and consider as possible model covariates all other variables in Table 1.6, except birth weight. The steps should include:
(i) a complete univariate analysis,
(ii) an appropriate selection of variables for a multivariate model (this should include scale identification for continuous covariates and assessment of the need for interactions),
(iii) an assessment of fit of the multivariate model,
(iv) preparation and presentation of a table containing the results of the final model (this table should contain point and interval estimates for all relevant odds ratios), and
(v) conclusions from the analysis.