Statistics

Question 1

What is type I error in statistics?

Answer 1

The error of concluding these is a difference from the null hypothesis when there is not (false positive)
Depends on p value (if set lower, more likely that the result is not due to chance alone)

Question 2

What is type II error is statistics

Answer 2

The error of concluding there is no difference from the null hypothesis when there is (false negative)
Depends on power (Power = 1-Type 2 error)

Question 3

Pre test probability

Answer 3

can assume to equal prevalence

Question 4

post test probability

Answer 4

approximates to PPV

Question 5

positive likelihood ratio

Answer 5

sens / (1-spec)

Question 6

negative likelihood ratio

Answer 6

1-sens / spec

Question 7

phase 1 trial

Answer 7

safety (dose range, side effects)

Question 8

phase 2 trial

Answer 8

efficacy (and more safety always)

Question 9

phase 3 trial

Answer 9

compare to standard therapy (monitor adverse effects and more safety)

Question 10

phase 4 trial

Answer 10

post marketing

effectiveness (and adverse events)

Question 11

prevalence

Answer 11

no effect on sensitivity, specificity or likelihood ratios
changes PPV, NPV
PPV higher with high prevalence, NPV lower
NPV higher with low prevalence, PPV lower

Question 12

You need increased numbers in a study if:

Answer 12

want to detect a smaller difference (e.g. 15% to 10%)
want to have higher power (e.g. from 80% to 90%), want to reject the null hypothesis with greater “proof” (0.05 to 0.01).