PRACTICAL Week 9 – Binary Logistic Regression


For this session, we will use a subset of the Health Survey for England.


We are interested in identifying the risk factors of hypertension. Hypertension is characterized by
elevated values of systolic or diastolic blood pressure. The cut-off points are 140mm Hg and 90 mm
Hg, respectively.


For this analysis we will focus only on the Systolic Blood Pressure. In our dataset it can be found
as the variable “omsysval”.



1.      Open the dataset Health_Survey_England.sav


2.      Paste all calculations and answers to the syntax file and save it regularly.



3.      Create dichotomous variable.

a.       Recode omsysval into new binary variables indicating whether the participant has
hypertension or not (hint: if omsysval>140 mm Hg). Before recoding, inspect the variable omsysval
for outliers and decide what to do with them.

                                                              i.      What is the prevalence of
hypertension?



4.      Test whether age predicts hypertension.

a.       Model fit

                                                              i.      Does the model fit the data
better than an empty model?

                                                            ii.      What are the Cox and Snell
R-square and Nagelkerke R-square?

                                                          iii.      What is the overall percentage
of cases that were correctly predicted by our model?

b.      Results

                                                              i.      What is the effect of age on
hypertension?

1.      What is the magnitude and significance of the slope?

2.      What is the magnitude and significance of the OR?

                                                            ii.      Write-up the results.


5.      Test whether sex predicts hypertension, after adjusting for the effect of age.

a.       Model fit

                                                              i.      Does the model fit the data
better than an empty model?

                                                            ii.      What are the Cox and Snell
R-square and Nagelkerke R-square?

                                                          iii.      How did these values change now
that we added another predictor?

                                                          iv.      What is the overall percentage
of cases that were correctly predicted by our model?

                                                            v.      Did the accuracy rate change
after adding another predictor?

b.      Results

                                                              i.      What is the effect of sex on
hypertension?

1.      What is the magnitude and significance of the slope?

2.      What is the magnitude and significance of the OR?

                                                            ii.      Write-up the results.


6.      Identify which of these common risk factors is associated with hypertension: age, sex, bmi,
cholesterol level, ethnicity, physical activity, smoking status and income.

a.       Build the model

                                                              i.      What can you say about the
model fit?

                                                            ii.      What about its accuracy?

b.      Results

                                                              i.      Which of these predictors is
associated with higher risk of elevated SBP?

                                                            ii.      Assuming that I am only
interested in assessing the effect of income on hypertension, how would you write up the result?


Submit your practical:

·         Save your syntax file to Homework Vaults