Example 1 Suppose we want to examine the association between unemployment days (x) and depression (y) by means of simple logistic regression analysis. Unemployment days are measured as the total number of days in unemployment during a year, and ranges from 0 to 365. Depression has the values 0=No and 1=Yes. Let us say that we get an OR that is 1.03. That would mean that we have a positive association: the higher the number of unemployment days, the higher the risk of depression.
Example 2 In another example, we may examine the association between intelligence scores (x) and drug use (y). Intelligence scores are measured by a series of tests that render various amounts of points, and ranges between 20 and 160 points. Drug use has the values 0=No and 1=Yes. Here, we get an OR of 0.91. We can thus conclude that the risk of using drugs decreases for every unit increase in intelligence scores.
Practical example
Dataset
StataData1.dta
Variable name
earlyret
Variable label
Early retirement (Age 50, Year 2020)
Value labels
0=No 1=Yes
Variable name
bmi
Variable label
N/A
Value labels
N/A
sum earlyret bmi if pop_logistic==1
logistic earlyret bmi if pop_logistic==1
When we look at the results for bmi, we see that the odds ratio (OR) is 1.00 or, more precisely, 1.005211. Thus, one unit increase in bmi does almost not change the odds of earlyret at all.
The association between bmi and earlyret is not statistically significant, as reflected in the p-value (0.60) and the 95% confidence intervals (0.99-1.03).
Summary There is a positive association between body mass index at age 15 and early retirement at age 50. The association is nonetheless very weak (OR=1.005) and statistically non-significant (95% CI=0.99-1.03).
