Get Answer: Part Interpreting Logistic Question Guide

Understanding this question requires applying core subject principles.

What This Question Is About

This question relates to part interpreting logistic and requires a structured academic response.

How to Approach This Question

Break the problem into smaller parts and analyze each logically.

Key Explanation

This topic involves part interpreting logistic. A strong answer should include explanation, application, and examples.

Original Question

Part 1. Interpreting Logistic Regression in Published Studies (30 points) Read the attached article by Bediako et al, titled “A comparison of birth outcomes among black, Hispanic, and black Hispanic women.” What is the objective of this study? State the exposure(s) and outcome(s) of interest. (10 points) State and interpret the unadjusted predicted probabilities for Black and White non-Hispanic women as predictors of low birth weight (from Table 2). (10 points) How do the adjusted predicted probabilities differ from the unadjusted predicted probabilities for Black and White non-Hispanic women as predictors of low birth weight (from Table 2)? (5 points) Based on the results of the interaction effects (Table 3), which factors had a differential effect on the association between race/ethnicity (black, white, Hispanic) and low birth weight? (5 points) Part 2. Logistic Regression, Binomial Regression and Log Binomial Regression (70 points) Using the birth weight dataset in SAS studio, we wish to investigate the effect of maternal pre-pregnancy weight in pounds (LWT) on infant low birth weight (LOW). Construct a univariate logistic regression model to evaluate the relationship between low birth weight and maternal pre-pregnancy weight. Are the conditions met for logistic regression? (Conditions listed on page Diez of Diez 3rd edition or Page 376 of Diez 4th edition). You do not need to create any graphs or run any analyses; just list the conditions and state how you would hypothetically assess if they are met). (6 points) What is the unadjusted odds ratios for the relationship between low birth weight and maternal pre-pregnancy weight? List the odds ratio and interpretation. (6 points) Based on this model, what is the probability of having a low birth weight baby for a mother who weighed 130 pounds prior to pregnancy? Show your calculation (See Example 9.31 on page 373 of Diez 4th Edition). (6 points) Model Selection Construct the following univariate (Model 1) and multivariable (Models 2-6) logistic regression models listed in the table below. Fill in the model fit statistics for each model. (18 points) Models Adjusted R2 Pre-Pregnant Weight (LWT) LWT + HT LWT + HT + PTL_yn LWT + HT + PTL_yn + Race LWT + HT + PTL_yn + Race + Age LWT + HT + PTL_yn + Race + Age + Smoke What do you conclude from the table? Which model represents the best fit (See Example 9.17 in Diez 4th edition)? Provide a table here of the best model (6 points) Run a backward selection on the full multivariable logistic regression model (Model 6). Provide a table here of the best model. What do you conclude? Does it agree with your answer in #9? How do they differ? (6 points) Which model do you choose to as the most predictive of low birth weight? List the odds ratios with 95% confidence intervals for each of the covariates in the model you choose. Interpret the odds ratios. (10 points, hint: don’t forget what’s in the model) Using the following model, run a binomial and log binomial model with the following specification: * Fill in the table for the listed risk differences and relative risks with the 95% confidence intervals for each of the covariates in the model.(Table=8 points) Covariate RD (95% CI) RR (95% CI) Pre-natal weight (LWT) Hypertension Race3 vs Race1 Smoke Interpret the RD and RR for LWT. (4 points, hint: don’t forget what’s in the model) *Notice the PTL_yn was left out in the binomial and log binomial models—see notes in the given code below. SAS Code for Logistic Regression *replace the first u59145102 with your user id*; proc import datafile=’/home/u59145102/my_shared_file_links/u59145102/lowbwt.csv’ dbms=csv out=lowbwt; run; data hw4; set work.lowbwt; if PTL=0 then PTL_yn=0; else if PTL ne . then PTL_yn=1; run; *Univariate logistic regression model for low birth weight; ods graphics on; proc logistic data=hw4 plots=all; model LOW(event=’1′) = LWT /rl rsquare; run; ods graphics off; proc logistic data=hw4 plots=all; model LOW(event=’1′) = LWT /rl rsquare; run; proc logistic data=hw4; model LOW(event=’1′) = LWT HT /rl rsquare; run; proc logistic data=hw4; model LOW(event=’1′) = LWT HT PTL_yn /rl rsquare; run; proc logistic data=hw4; class Race(ref=’1′) /param=ref; model LOW(event=’1′) = LWT HT PTL_yn Race /rl rsquare; run; proc logistic data=hw4; class Race(ref=’1′) /param=ref; model LOW(event=’1′) = LWT HT PTL_yn Race Age /rl rsquare; run; proc logistic data=hw4; class Race(ref=’1′) /param=ref; model LOW(event=’1′) = LWT HT PTL_yn Race Age Smoke /rl rsquare; run; *Backward selection; proc logistic data=hw4; class Race(ref=’1′) /param=ref; model LOW(event=’1′) = LWT HT PTL_yn Race Age Smoke /rl selection=backward; run; *Alternative code for logistic models using proc genmod, have to do a bit more work to get your odds ratios; *have to specify the link and the distribution, /link=logit dist=bin, and ORs are not automatically generated, so have to specify them with “estimate”; /* proc genmod data=hw4; class Race(ref=’1′) /param=ref; */ /* model LOW(event=’1′) = LWT HT PTL_yn Race Age Smoke /link=logit dist=bin; */ /* estimate “ORlwt” LWT 1/exp; */ /* estimate “ORht” HT 1/exp; */ /* estimate “ORptl” PTL_yn 1/exp; */ /* estimate “ORr2” RACE 1 0/exp; */ /* estimate “ORr3” RACE 0 1/exp; */ /* estimate “ORage” AGE 1/exp; */ /* estimate “ORsmk” Smoke 1/exp; */ /* run; */ *code for log binomial models (relative risk); proc genmod data=hw4; class Race(ref=’1′) /param=ref; model LOW(event=’1′) = LWT HT Race Age Smoke /link=log dist=binomial; estimate “RRlwt” LWT 1/exp; estimate “RRht” HT 1/exp; estimate “RRr2” RACE 1 0/exp; estimate “RRr3” RACE 0 1/exp; estimate “RRage” AGE 1/exp; estimate “RRsmk” Smoke 1/exp; run; *code for binomial models (risk difference); /* proc genmod data=hw4 plots=all; model LOW(event=’1′) = LWT /link=identity dist=binomial; run; *crude model; */ proc genmod data=hw4; class Race(ref=’1′) /param=ref; model LOW(event=’1′) = LWT HT Race Age Smoke /link=identity dist=binomial; estimate “RDlwt” LWT 1; estimate “RDht” HT 1; estimate “RDr2” RACE 1 0; estimate “RDr3” RACE 0 1; estimate “RDage” AGE 1; estimate “RDsmk” Smoke 1; run;

******CLICK ORDER NOW BELOW AND OUR WRITERS WILL WRITE AN ANSWER TO THIS ASSIGNMENT OR ANY OTHER ASSIGNMENT, DISCUSSION, ESSAY, HOMEWORK OR QUESTION YOU MAY HAVE. OUR PAPERS ARE PLAGIARISM FREE*******."