Biostats SPSS (The Effect of Weights)

PART2

Week 11 Step by Step Application Guide 11.2

Effect of Weights

Problem 2. Logistic Regression

a. Use SPSS to run a logistic regression model with Q22a. “Have you ever looked online for -- Information about a specific disease or medical problem?” as your dependent variable (Note that there are 4 levels of responses possible, but only 2 are actually used in the responses so you can state the dependent variable is a binomial and use binary logistic regression) and Sex as the independent variable.

Step 1. Analyze ( Regression ( Binary Logistic

Step 2. Move Q22a Have you ever looked online… to Dependent. Move Sex to Covariates.

Step 3. Click Categorical. Move sex to Categorical Covariates. Move Reference Category to First. Click continue.

Step 4. Select Options. Check CI for exp(B) 95%. Click Continue. Click OK.

b. Use backwards stepwise regression to add Receduc to the model as a potential confounder.

Step 6: Select Analyze ( Regression ( Binary Logistic.

Step 7: Move Receduc variable into Covariates with Sex.

Step 3: Click Categorical. Move Receduc from Covariates to Categorical Covariates. Change Reference Category to First. Click Change. Click Continue.

Step 4.. In Logistic Regression window, click on the down arrow in the box for Method. Select Backward: LR. Click OK.

c. How does the relationship between Q22a and Sex change with the addition of Receduc? Include a discussion of Odds Ratios and the Model Summary in your answer. Would you consider Receduc a confounder? Is it worth keeping it in the model even if it does not confound the relationship between Q22a and Sex in this sample?

Note that for the dependent variable (Q22a: looked up online), 0 = yes, 1 = no. This is critical for interpreting the results.

Dependent Variable Encoding
Original Value	Internal Value
Yes, have done this	0
No, have not done this	1

Variables in the Equation
	B	S.E.	Wald	df	Sig.	Exp(B)	95% C.I.for EXP(B)
							Lower	Upper
Step 1a	sex(1)	-.625	.095	42.961	1	.000	.535	.444	.645
	Constant	-.310	.070	19.817	1	.000	.734
a. Variable(s) entered on step 1: sex.

Be sure to discuss what the OR and CI for sex means in this table.

Method = Logistic Regression with Sex only

Model Summary
Step	-2 Log likelihood	Cox & Snell R Square	Nagelkerke R Square
1	2532.536a	.021	.029
a. Estimation terminated at iteration number 3 because parameter estimates changed by less than .001.
Variables in the Equation
	B	S.E.	Wald	df	Sig.	Exp(B)	95% C.I.for EXP(B)
							Lower	Upper
Step 1a	sex(1)	-.625	.095	42.961	1	.000	.535	.444	.645
	Constant	-.310	.070	19.817	1	.000	.734
Method = Backward Stepwise (Likelihood Ratio) with Sex and Receduc Model Summary Step -2 Log likelihood Cox & Snell R Square Nagelkerke R Square 1 2291.692a .122 .169 a. Estimation terminated at iteration number 5 because parameter estimates changed by less than .001.

Variables in the Equation
	B	S.E.	Wald	df	Sig.	Exp(B)	95% C.I.for EXP(B)
							Lower	Upper
Step 1a	sex(1)	-.633	.102	38.623	1	.000	.531	.435	.648
	receduc			146.126	3	.000
	receduc(1)	-2.797	.472	35.083	1	.000	.061	.024	.154
	receduc(2)	-3.587	.474	57.377	1	.000	.028	.011	.070
	receduc(3)	-3.940	.472	69.658	1	.000	.019	.008	.049
	Constant	3.053	.469	42.315	1	.000	21.172
a. Variable(s) entered on step 1: sex, receduc.

Receduc(1)= HS grad; Receduc(2)=Some college; Receduc (3)=College+

Be sure to include a discussion of the changes in the model with the addition of education. How is the OR and 95% CI for Sex changed? Is education significantly associated with looking at information online based on the OR and CI for the differing levels (Dummy variables receduc 1, 2, and 3?