Multiple linear regressions and multiple logistic regressions are similar in that they allow you to adjust for potential confounders when testing the relationship between your dependent and independent variables. SPSS can be used to conduct both of these tests. It is important to note, however, that the interpretation of the output with logistic regression is very different from that with linear regression.
In the following two problems the analysis has been conducted for you using SPSS.
Problem 1:First Problem Output
Interpret the results by answering the following questions:
1)Was the relationship between sex and FEV significant?
2)What percentage of the variability of FEV was explained by sex? (Use the coefficient of determination.)
3)Was the relationship of age and FEV significant?
4)Was age a confounder of the relationship between sex and FEV?
5)What percentage of the variability of FEV was explained by sex and age? (use coefficient of determination)
Problem 2:
A case-control study was conducted on patients undergoing extensive cardiac surgery. Cases were given pre-operative medication (study medication) to increase their chance of survival. The outcome of interest is "alive discharge from the hospital" (survival). The relationship between assignment as a case or a control, and survival is tested using logistic regression. See the document below for this output (Second Problem Output, first part).
Then, the relationship between assignment as a case or a control, and survival is tested again, this time adjusting for the presence of sepsis (a whole body response to infection) after surgery using multiple logistic regression. See the document below for this output (Second Problem Output, second part):
Problem 3:Second Problem Output
Interpret the results by answering the following questions:
1) Was the relationship between case/control and survival significant?
2) Was the relationship of sepsis and survival significant?
3) Was sepsis a confounder of the relationship between case/control and survival?
4) What would the researchers have concluded about the study medication if they had ignored the effect of sepsis on survival in their analysis?