Question

Consider a new (and major) issue in biomedical and other types of research: datasets on many conditions and treatments are now extremely large, due to the availability of electronic records.

Suppose you are comparing outcomes in a study that compares weight gain under two widely used diet plans for patients who are receiving chemotherapy, where weight loss is common.

a) What will the large samples do to the standard errors?

b) How will these standard errors impact t-tests used to compare weight in the two groups?

c) Why is this an issue?

Answer 1

a) A large set of samples tends to reduce standard errors. Because of the more 'N' in the denominator, the lesser would be the error occurred.

b) t-test also will have lowered Standard deviation and would give a lower p-value if the null hypothesis is rejected or it might give higher p-value (near to 1) if the null hypothesis is accepted (where the null hypothesis is that the two test group's outcomes have no difference)

c) The issue is to refine the data and discard the outliers or evitable mistakes, but fishing out the mistake from big data set is really challenge and issue at hand.