The Glucose2 dataframe has three columns, one for each of three pregnancies.
For the first two columns, test1 and test2, test the hypothesis of no change, using the paired t-test. Do the differences appear to be approximately normally distributed?
For bootstrapping a paired sample, we need to select the subjects, not individual values, since the values come paired by subject. For the Glucose2 dataset, there are 52 subjects. To select a single bootstrap sample of size 52 from the 52 subjects, use:
n <- sample(1:52,replace=T)Then, for example, test1[n] and test2[n] are corresponding pairs of observations, and mean(test2[n]-test1[n]) returns the mean difference for the bootstrap sample. Alternatively, one could make a dataset containing the differences (test1-test2), and resample that dataset.
Construct a 95% CI using the bootstrap, based on at least 20000 bootstrap trials.
Do the t-test and bootstrap analyses agree? If not, on which should we base our conclusions and why? What should we conclude?