Statistics Flashcards

1
Q

PPV and NPV changes with prevalence

A

PPV will rise with higher prevalence

NPV will rise with lower prevalence

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

Positive and negative likelihood ratio equation

A

Positive: sens/ (1-spec)
Negative: (1-sens)/ spec

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

Clinical trials phases 0-4

A

Phase 0: prelim PK/PD info
Phase 1: healthy volunteers for safety and toxicity, PK, PD
Phase 2: larger to look at dose and efficacy
Phase 3: RCT for efficacy (cf good standard)
Phase 4: post marketing survellience for safety vigilance

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

Adoption vs twin studies

A

Adoption to look at genetic influences in same environment

Twin studies to review affects of genes (mono vs diz)

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

Cohort studies down side

A

Prone to loss to follow up

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

Pre registration of trials prevents?

A

Publication bias

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

Type 1 and 2 error

A

1 error concluding there is different when one doesn’t exist (p value) “false positive”
2 error concluding there is no difference when there is one (power) “false negative”

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

SD 1 standard deviation and 2 standard deviation (percentage)

A

1: 68% (16% above and 16% below)
2: 95%
3: 99.7%

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

What is variance?

A

SD squared

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

Intention to treat important why

A

Underestimated true effect

Maintains randomisation to overcome selection bias

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

Pre test odds take into account

A

Patient characteristics and prevalence

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

Multivariate analysis two groups

A
Multiple regression (continuous variable)
Logistic regression (dichotomous)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Size of box on forest plot

A

Related to quality and size of study

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

Your region has a birth rate of about 50,000 babies per year. About 50 preschool children are diagnosed
each year with “crumblitis”, a rare gastrointestinal disorder, the cause of which is unknown. You hypothesise
that this condition is related to maternal ingestion of orange rhubarb during pregnancy.
Which one of the following is the most appropriate research design to test your hypothesis?

A

Case control

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

SIDS percent of deaths

A

7%

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

Severe disability rate

A

1/25

17
Q

Children with chronic condition

A

37%

18
Q

2003 (a) QUESTION 53
What is the greatest strength of the case-control study design?
A. Controls for bias.
B. Controls for confounding.
C. Feasible with small or moderate numbers.
D. Multiple endpoints can be assessed.
E. Multiple risk factors can be assessed.

A

Case control is comparing characteristics of those with disease/ outcome to a control
group without
This is good for conditions of low prevalence where it would take a very long time to
collect enough numbers in a RCT
In general determinesassociatiations rather than causation (esp if exposure occurred in
past)
ANSWER = C

19
Q

P value ‘significance’

A

0.05
The smaller the p value the smaller the chance that the null hypothesis is true (null meaning no difference between the groups and most studies are trying to reject this)

20
Q

High correlation if r =

A

0.6-0.8
Rsquared is the percentage of data related to intervention with a correlation change
Correlation is strength between two variables
Regression quantifies it (straight line through scatter)

21
Q

What is kappa

A

Level of agreement when 1 is perfect agreement (0.5 or more is good)

22
Q

Incidence vs prevalence

A

Incidence is NEW diagnoses

Prevelance is overall

23
Q

Chi squared is

A

Looking at if there is a significant difference between two categorical variables