Stats, study design

Question 1

What percentage of data included in +/- 1 standard deviation from the mean?

Answer 1

68.2%

Question 2

How do you calculate SD from variance?

Answer 2

SD = square route of variance

Question 3

Group of 100 patients with knee injuries (“cases”)
Matched for age and sex to 100 patients without knee injuries (“controls”)
In the cases – 40 were skiers, 60 were non-skiers
In the controls – 20 were skiers, 80 were non skiers.
What are the ORs for the control and case groups?

Answer 3

Odds of a case being a skier? 40/60 = 0.66
Odds of being a skier? 20/80 = 0.25
OR = 0.66/0.25 = 2.64

Question 4

How would you calculate the ARR of an intervention in a cohort study evaluating the effect of a new drug compared to placebo?

Answer 4

(absolute risk reduction) difference between the event rate in the intervention group and control group
ARR = risk in exposed group - risk in control group

Question 5

How would you calculate RRR?

Answer 5

(relative risk reduction) the proportion by which the intervention REDUCES the event rate
= ARR/control (placebo) no improvement event rate

Question 6

How would you calculate NNT for an intervention?

Answer 6

NNT = 100/ARR

Question 7

How would you calculate sensitivity from a 2-way table?

Answer 7

Senstivity = A / A+C

Question 8

How would you calculate specificity from a 2-way table?

Answer 8

Specificity = D / B+D

Question 9

How would you calculate PPV from a 2-way table? (If the result is positive, what is the likelihood the patient will have the condition?)

Answer 9

PPV = A / A+B

Question 10

How would you calculate NPV from a 2-way table? (If the result is negative, what is the likelihood the patient is healthy?)

Answer 10

NPV = D / C+D

Question 11

Define LR+

How would you calculate this?

Answer 11

LR + is the multiplier for how much more likely the patient is to have the condition if the test result is positive

LR+ = sensitivity / (1-specificity)

Question 12

Define LR-

How would you calculate this?

Answer 12

LR – is the multiplier for how much the risk of having the condition has decreased if the test is negative
LR- = (1-specificity) / sensitivity

Question 13

What is the I2 statistic for heterogeneity?

Answer 13

Measure of heterogeneity of studies included in meta-analysis or systmatic reviews
If I2 = 0% - homogenous
If I2 > 50% - high level of heterogeneity

Question 14

What is the kappa statistic used to describe?

Answer 14

Levels of agreement in ordinal data
Eg levels of agreement of 2 hospitals viewing cervical cytology samples
0 = no agreement
1 = full agreement

Question 15

When would the intra-class correlation coefficient (ICC) be used?

Answer 15

When looking at levels of agreement in continuous data eg 2 different blood glucose measuring devices

Question 16

What letter is the correlation coefficient denoted by?

Answer 16

R

Question 17

How is r squared used in interpretting correlation?

Answer 17

tells you how much the variation in one value is related to variation in the other value
eg 50% of variation in aortic systolic pressure is accounted by left ventricular end systolic area