STAT stuff

Question 1

What is prevalence

Answer 1

prescence of condition in pop at any given time
—- True positive —- TP
— False +: FP
False negative - FN
true negative - TN

— prevalence : TP+FN/ Total #

Question 2

What is sensitivity

Answer 2

prob of getting a correct result (+) given you have the disease
- TP/TP=FN

—- chance you get + if you are +

Question 3

What is specificity

Answer 3

prob of getting a correct result (-) when disease is absent
- TN /FP+TN

chance you get - given you are -

Question 5

What is false +

Answer 5

prob of getting incorrect result (+) given you don’t have disease

FP/FP+TN OR 1 - specificity

** percentage of those who don’t have disease that get +

Question 6

What is false -

Answer 6

1- sensitivity
- prob of getting an incorrect result (-) given you have the disease

FN /TP+FN

percentage with disease who get -

Question 7

Positive Predictive Value

Answer 7

prob of actually having the disease given you get a + result

— if get + result: how likely is it that I am actually +

TP/TP+FP

Question 8

Negative Predictive Value

Answer 8

if - test: how likely is it that I am actually -

TN/FN+TN

Question 9

Low sensitivity will cause an increase in false ____ rate

Answer 9

False - rate
—- if sensitivity low; if we are -, not good at ruling out that - is ACTUALLY -

*low sensitivity: disease more likely to not be detected when present

Question 10

Low specificity will cause an increase in false ___ rate

Answer 10

false +
— low specificity: if +, less likely you are actually + (not good at ruling in)

Question 11

High specificity is good at ———-

Answer 11

ruling in (if + —- you +)

Question 12

High sensitivity is good at ____

Answer 12

ruling out

Question 13

T or F: predictive values are not dependent on prevalence

Answer 13

F- highly dependent
—- if do tests in different pops; will get roughly same sensitivity + specificity BUT different +/- predictive values

Question 14

What are likelihood ratios

Answer 14

prob of a test result for a person w/diease divided by prob of the test result for a person w/out disease
—- used in place of predictive values when looking at different cohorts

Question 15

Formula for LR

Answer 15

sensitivity/1-specificity
aka T+ rate/false + rate

Question 16

T or F: we want LR to be a super number less than 10

Answer 16

F- larger is better + aim for > 10

Question 17

What does changing cutoff values due to sensitivity, specificity +/- predictive value

Answer 17

change them

• low cutoff (lower value —- will give + result): higher sensitivity (if - : you - / low FN ) but low specificity (will get mor false +)

high cutof (need high value for +): low sensitivity (more FN) + high specificity (low FP)

Question 18

What is LR+

Answer 18

likelihood ratio of + test result
—- prob of + result for person with disease/ prob of + result for person without disease

** interpretation of #: that is the number of times you are more likely to get + result if you have disease vs if have no disease

** higher is better — better diagnostic test

Question 19

What is a ROC curve

Answer 19

Receiver Operating Charactersitic Curve
— graph of false + rate (X axis) vs true + rate (y)

—— graph of sensitivity vs F+ rate

used to determine good cutoff values

Question 20

What is desirable in a ROC curve

Answer 20

AUC closer to 1 + want curve closer to Y axis (indicate higher T + rate and low False + rate)

—= best combo of sensitivity + specificity (maximize them)