Interpretations Flashcards

1
Q

odds ratio interpretation

A

> assesses how a 1 unit increase in predictor variable affects likelihood (odds) of response
odds > 1 = likelihood increases as predictor increases
odds < 1 = likelihood decreases as predictor increases

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Hazard ratio interpretation

A

> HR = 1: no difference in risk between groups
HR > 1: the event is x times more likely to occur to this group.
HR < 1: the event is x times less likely to occur in this group compared to….

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

per-protocol analysis definition

A

> only participants who fully adhered to study protocol are included in final analysis
dropouts are excluded
not censored; fully removed

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

intention to treat analysis definition

A

> includes all participants in analysis as per their initial group regardless of if they followed study protocol
maintains randomisation
reflects real-world treatment effects

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

data cleaning rules

A

> believe a prior question over a subsequent
believe and answered question over a missed one

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Data checking steps

A

> check inclusion criteria
check variables for false/incorrect values

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Difference between kruskal-wallis and mann-whitney tests for continuous data

A

> both for non-parametric
mann-whitney tends to be for differences between 2 groups (t-test)
kruskal-wallis is for differences in 3 or more groups (ANOVA)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Data type for logistic regression

A

binary outcome; to model with covariates

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Data type for linear regression

A

continuous outcome; model with covariates

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Data type for poisson regression

A

outcome is count data; model with covariates.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

time-to-event analysis

A

> kaplan-meier
cox proportional hazards
logrank test

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Analysis for categorical data

A

> chi - squared
fisher’s exact test
mann-whitney (if ordinal)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

How does randomisation conclude causality?

A

because groups are balanced on any other differences and there is no subjective choice involved

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Why are composite outcomes used?

A

Used to increase power when the event rates for the individual events are small.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Composite outcome meaning

A

there are two separate endpoints combined, both of which measure the same intervention effect.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

randomisation ensures…

17
Q

hypotheses for proportion studies

A

proportions equal/proportions not the same

18
Q

What is cox proportional hazards model testing?

A

null hypothesis that the hazard ratio is 1

19
Q

what is logistic regression model testing?

A

null hypothesis that odds ratio = 1

20
Q

Test to use for large sample, numerical data

A

Z test for difference of two means (same as t but for large samples)