06 Sood Lecture 2

Question 1

What does skewed data imply?

Answer 1

That parametric hypothesis testing (t-test) does not work in small samples (less than 150)

Question 2

What needs to be done when there is missing cost data?

Answer 2

Solution to these problems involves imputing missing values based on similar patient who has non-missing data. This will only work if those with missing data are a random sample

Question 3

What are the steps in calculating confidence intervals using non-parametric bootstrap?

Answer 3

Draw a random sample with replacement from observations of treatment group. Draw a random sample with replacement from observations of control group. Calculate ICER for given “bootstrap” sample. Repeat steps 1-3 1000 times. Rank ICERs. CI of ICER is ICER of 25th ranked to 975th ranked observation