Author Question: A researcher decides to study how many college freshmen with continuous enrollment graduate within ... (Read 119 times)

Beheh · **on:** Jul 8, 2018

A researcher decides to study how many college freshmen with continuous enrollment graduate within four years, four-and-a-half years, five years, five-and-a-half years, and six years,

and to analyze this according to what their majors at graduation are, as well as other demographic variables. Why would a Cox proportional hazards regression analysis be suitable for this research? (Select all that apply.)
a. Demographic variables could affect time until graduation.
b. The data of people who do not graduate at all may be informative.
c. Attendance at college can present hazard.
d. A predictor of the dependent variable is time-related.
e. Only people who actually graduate by the end of twelve semesters will have their data analyzed.

Question 2

Why does a researcher decide to use logistic regression instead of linear regression to calculate survival odds after suffering an out-of-hospital heart attack, in persons of various ages, genders, and cardiac diagnoses? (Select all that apply.)

a. Survival versus non-survival is a dichotomous variable.
b. Having a heart attack is a variable that has many possible outcomes.
c. Linear regression is less logical than is logistic regression.
d. Age is a ratio-level variable.
e. Gender is a nominal-level variable.
f. Cardiac diagnosis is a nominal-level variable.

jomama · **Reply #1 on:** Jul 8, 2018

Answer to Question 1

ANS: A, B, D
The purpose of a regression analysis is to predict or explain as much of the variance in the value of a dependent variable as possible. The major difference between using Cox regression as opposed to linear regression is the ability of survival analysis to handle cases where survival time is unknown. Whereas logistic regression yields odds ratios for each predictor to represent the relationship between that predictor and y, Cox regression yields hazard ratios.

Answer to Question 2

ANS: A, D, E, F
Logistic regression replaces linear regression when the researcher wants to test a predictor or predictors of a dichotomous dependent variable. Logistic regression can have continuous predictors or nominal predictors or a combination of the two, with no assumptions regarding normality of the distribution.