End of Module C1 Quiz

Based on the below two-by-two table, choose all that are correct.

A cohort study on smoking and dementia reported a relative risk of dementia = 1.03 (95% CI 0.75-1.41) for smokers compared with non-smokers. How would you interpret this?

In a dataset of 3000 male current drinkers, a chi-square test for association was performed between current smoking (1=yes vs. 0=no) and alcohol intake (recorded by a categorical variable wkcat; 1: 1-139 g/week, 2: 140-279 g/week, 3: 280-419 g/week, 4: 420+ g/week). The following output was obtained. What can you conclude based on the chi-square test result?

When would you want to perform a chi-square test for linear trend instead of (or in addition to) a chi-square test for association?

Questions 5-9 are based on ananalyses conducted in a fictitious study of 20000 Chinese men to investigate the risk factors for prevalent chronic obstructive pulmonary disease (COPD) (recorded by the binary variable has_copd ).

A logistic regression model was run to investigate the relationship between prevalent COPD and age (continuous variable, in years). How would you interpret the output?

[Continuing from the previous question]

Another logistic regression model was run to investigate the relationship between COPD and education groups (recorded by a categorical variable called education, coded from low to high level: 1= no formal school; 2= primary school; 3= middle or high school; 4= technical school/college or above). How would you interpret the results?

[Continuing from the previous question]

Instead of a categorical variable, this time the education variable was input as an ordinal variable in the logistic regression. You obtained a single OR of prevalent COPD associated with the ordinal variable education, which is 0.54 (95% CI: 0.51-0.58) with p<0.001. Choose all that are correct.

After running separate univariable logistic regression models on prevalent COPD with age, education, ever-alcohol drinking (recorded by a binary variable evralc), and ever-smoking (recorded by a binary variable evrsmk), all of there variables were significantly associated with COPD. However, you wondered if confounding could explain some of these unvariable associations. You therefore ran a multivariable logistic regression model including all these variables. How would you interpret the results?

You wanted to investigate if there is potential interaction between smoking and age on prevalent COPD. You fitted the below multivariable model with an interaction term between ever-smoking (binary variable evrsmk: 0= never-smokers, 1= ever-smokers) and age groups (binary variable agebin : 0= below 60 years, 1 = 60 years or above). Based on these output, choose all that are correct.

(modified from Discovering Statistics Using IBM SPSS Statistics, by Andy Field)

When analysing continuous predictor variables in logistic regression, we assume a: