Stats

Question 1

What is the p value?

Answer 1

Probability that your trial will say a therapy works given that it actually doesn’t

OUTCOME | NULL

Question 2

What is sensitivity and how do you calculate it?

Answer 2

SNOUT - Sensitive test is good at ruling out conditions when the test result is negative [Proportion of those who test positive among those who actually have the disease)

True positive/(true positive + false negative)

Question 3

What is specificity and how do you calculate it?

Answer 3

SPIN - Specific test is good at ruling in conditions when the test result is positive [Proportion of those who test negative among those who those who don’t actually have the disease]

True negative/(true negative + false positive)

Question 4

What is PPV and how do you calculate it?

Answer 4

Probability that following a positive result, the individual will truly have the disease

True positive/(true positive + false positive)

Question 5

What is NPV and how do you calculate it?

Answer 5

Probably that following a negative result, the individual truly won’t have the disease

True negative/(true negative + false negative)

Question 6

How do you calculate the following: AR, ARR, RR, RRR, NNT?

Answer 6

AR: no. of events (good or bad) in the group (control or treatment) divided by number of people in the group

ARR: ARC - ART

RR: ART/ARC

RRR: 1 - RR

NNT: 1/ARR

Question 7

How do you work out RR and OR from risk tables?

Answer 7

RR = (A/A+B) / (C/C+D)

OR = (A/C / (B/D)

Question 8

How do you work out power calculations for binary outcomes?

Answer 8

30 x W^2 x R

R = how rare your event is (e.g 1 in 10 deaths, R = 10)
W = how many events your therapy is trying to prevent in the hypothesis (e.g 1 in 4 deaths, W=4)

Question 9

How do you work out power calculations for continuous outcomes?

Answer 9

30 x (S/D)^2

S = Standard deviation of difference

D = minimum effect size if there is a difference in means (e.g hypothesis states change in EF of 4%, D =4)