STATS

Question 1

What is the hierarchy of study designs?

Answer 1

1. Systematic review and meta-analysis
2. RCT
3. Cohort studies
4. Case-control studies
5. Cross-sectional studies
6. Case series/expert opinion

Question 2

Which study type involves two or more groups followed over time and compared with respect to an outcome?

Answer 2

(prospective) cohort study

Question 3

Which study type involves groups of patients with or without an outcome that are compared for past exposures or conditions?

Answer 3

case-control study

Question 4

Which study type involves an outcome and exposure being measured at the same time?

Answer 4

cross-sectional study

Question 5

Which study can determine incidence?

Answer 5

cohort study

Question 6

Which study can determine prevalence?

Answer 6

cross-sectional study

Question 7

Which study is useful in rare diseases or outcomes?

Answer 7

case-control study

Question 8

A cohort study can determine what?

Answer 8

incidence (can distinguish cause and effect)

Question 9

A cross-sectional study can determine what?

Answer 9

prevalence (cannot distinguish cause and effect)

Question 10

What study is required to establish cause and effect?

Answer 10

cohort study (or higher)

Question 11

Which studies are observational studies?

Answer 11

cohort, case-control, and cross-sectional

Question 12

If a confidence interval includes what number, the possibility of no effect cannot be ruled out?

Answer 12

0

Question 13

What statistical test is the plausibility of the null hypothesis?

Answer 13

p-value

Question 14

What is the type I error rate?

Answer 14

false positive rate or alpha (p-value)

Question 15

What is the false positive rate?

Answer 15

alpha (p-value) or type I error rate (1-specificity)

Question 16

What is the false negative rate?

Answer 16

Beta or type II error (1-sensitivity)

Question 17

What is a type II error?

Answer 17

Beta or false negative rate

Question 18

Power =

Answer 18

Power = 1-B

Question 19

Which variables are numerical?

Answer 19

continuous variables

Question 20

Which variables are non-numerical?

Answer 20

categorical variables

Question 21

Which statistical test assess the average of continuous variables?

Answer 21

T test

Question 22

Which statistical test assess the distribution of continuous variables?

Answer 22

Wilcoxon rank-sum test or Mann-Whitney U test

Question 23

Which statistical test assesses if categorical variables are independent (no rare variables)?

Answer 23

Chi-squared test

Question 24

Which statistical test assesses if categorical variables are independent (yes rare variables)?

Answer 24

Fisher’s exact test

Question 25

In a 2x2 table, what is the EER (experimental event rate)?

Answer 25

EER = A/(A+B)

Question 26

In a 2x2 table, what is the CER (control event rate)?

Answer 26

CER = C/(C+D)

Question 27

How is a 2x2 table set up?

Answer 27

Top (outcome or disease): Present, Absent; Left (treatment or test): Treatment, Control or Positive, Negative

Question 28

In a 2x2 table, what is RR (relative risk)?

Answer 28

RR = EER/CER

Question 29

In a 2x2 table, what is RRR (relative risk reduction)?

Answer 29

RRR = (CER-ERR)/CER or = 1-RR

Question 30

In a 2x2 table, what is the ARR (absolute risk reduction)?

Answer 30

ARR = CER-EER

Question 31

In a 2x2 table, what is the NNT?

Answer 31

NNT = 1/ARR

Question 32

What is a methodology in which the relationship between two or more variables is estimated?

Answer 32

Regression analysis

Question 33

What is a model in which the odds of an outcome are associated with one or more variables?

Answer 33

Logistic regression

Question 34

In a 2x2 table, what is OR (odds ratio)?

Answer 34

(A/B)/(C/D) or (A+D)/(B+C)

Question 35

To assess how well a logistic regression model is able to predict its outcome, what can you use?

Answer 35

ROC curve

Question 36

What is the true positive rate?

Answer 36

sensitivity

Question 37

What is the true negative rate?

Answer 37

specificity

Question 38

What is the false positive rate?

Answer 38

1-Specificity

Question 39

In a 2x2 table, what is sensitivity?

Answer 39

sensitivity = A/(A+C) = TP/(TP +FN)

Question 40

In a 2x2 table, what is specificity?

Answer 40

specificity = D/(B+D) = TN/(FP+TN)

Question 41

In a 2x2 table, what is PPV?

Answer 41

PPV = A/(A+B) = TP/(TP+FP)

Question 42

In a 2x2 table, what is NPV?

Answer 42

NPV = D/(C+D) = TN/(FN+TN)

Question 43

What is the y-axis of a ROC?

Answer 43

sensitivity (true positive rate)

Question 44

What is the x-axis of a ROC?

Answer 44

1-specificity (false positive rate)

Question 45

What is a model in which the outcome (continuous variable) is related to one or more variables?

Answer 45

linear regression

Question 45

What are the results of a logistic regression?

Answer 45

odds ratio, CIs, p-values

Question 45

What model can assess the relationship of one or more variables to the occurence of an event? ("survival analysis" or "time-to-event analysis")

Answer 45

Cox proportional hazards regression

Question 46

What are the results from a linear regression?

Answer 46

estimated effect size, CIs, p-values

Question 46

What are the results of a Cox proportional hazards regression?

Answer 46

hazard ratio, CIs, p-values

Question 47

If using continuous variables in any regression model, the odds ratio, effect size or hazard ratio show the magnitude of association for each increase in what?

Answer 47

one unit (eg, estimated effect is -0.5 for a linear regression of Hgb and weight; for each 1kg of weight, Hgb decreases by 0.5)

Question 48

For a continuous variable regression analysis, how to calculate an odds ratio increasing by 5 units (instead of 1 unit)?

Answer 48

(OR)^5

Question 49

What is it called when other variables are included as predictors in a regression model?

Answer 49

controlling (or adjusting) for those variables

Question 50

What is the correction used in a multiple comparison study to limit higher "false positive" rates?

Answer 50

Bonferroni correction

Question 51

How to perform a Bonferroni correction?

Answer 51

Take initial p-value (0.05) and divide by total number of tests (outcomes x variables) to get new p-value

Question 52

When is an ANOVA used?

Answer 52

instead of a T-test (normal distributions) when 3+ groups