1
Q
Define mean
A
- average of all scores
2
Q
Define median
A
- after ranking scores, the score in the middle
- separates the higher half from the lower half
- helps identify major outliers
3
Q
Define mode
A
- score that occurs the most frequently
4
Q
Explain regression toward the mean
A
- if you take averages of the same studies over and over and over you should get a regression closer to the actual mean
5
Q
Independent variable
A
- manipulated via researcher
6
Q
Dependent variable
A
- what is measured
7
Q
Define validity
A
- truthfulness, meaningfulness, and or accuracy of study results
8
Q
2 categories of validity
A
- external validity
- internal validity
9
Q
Define external validity
A
- generalizability of results
10
Q
Define internal validity
A
- controlled by the study design
- ie blinding
11
Q
What are the 5 types of validity
A
- face
- content
- concurrent
- predictive
- construct
12
Q
Define face validity
A
- does a specific measure actually measure what it was designed to measure
13
Q
Define content validity
A
- does the measure represent all constructs of the measure that it is intended to
14
Q
Define concurrent validity
A
- correlates with gold standard
- how well a new assessment aligns with gold standard
15
Q
Define predictive validity
A
- test can be used to predict a future score or outcomes
16
Q
Define construct validity
A
- how well measure captures defined entity
- how well measure captures the theoretical concept it’s meant to measure
17
Q
How is validity typically measured
A
- correlation (closer to 1 is better on a scale of 0-1)
18
Q
Interval and ratio data use what statistical analysis
A
pearson
19
Q
ordinal uses what statistical analysis
A
spearman rank
20
Q
nominal uses what statistical analysis
A
phi
21
Q
define reliability
A
- consistency of a specific measure
- able to produce consistent repeated measures
22
Q
what are the 2 types of reliability
A
- intrarater
- interrater
23
Q
Define intrarater reliability
A
- consistency and reliability of a single rater
24
Q
Define interrater reliability
A
- consistency and reliability of multiple raters
25
Measures of reliability are expressed as what
- intraclass correlation coefficient (ICC)
26
ICC requires what type of data
nominal or ordinal
27
What is considered a good ICC
greater than 0.75
28
What is moderate ICC
0.51--.75
29
What is low ICC
less than 0.50
30
What is agreement
- do the measure and say whether it is negative or positive
- percent agreement
- the most basic
31
What is the kappa statistic
- considers the agreement by chance to help factor chance out
32
almost perfect agreement between raters is what
0.81-1.0
33
substantial agreement between raters is what
0.61-0.80
34
Moderate agreement between raters is
0.41-0.60
35
fair agreement between raters is
0.21-0.40
36
Slight agreement between raters is
0.01-0.20
37
poor/equal to change agreement between raters is
less than or equal to 0
38
define precision
- absolute reliability
39
SEM is derived from what
- ICC and standard deviation
39
What is precision expressed as
- standard error of measurement (SEM) which provides insight into meaningful changes
40
Define minimal detectable change (MDC)
- smallest amount of change an instrument can accurately measure
- changes must exceed MDC to be beyond measurement error (if below change is too small to say real change occurred)
- no context, just that change is beyond error
41
Define minimal clinically important difference (MCID)
- smallest difference that clinicians and patients would care about (what is the change that's actually meaningful clinically)
42
What does MCID tell us
- smallest change needed for it to be clinically significant
43
Define ceiling effect
- instrument does not register a further increase in score for high scoring individuals
- scores all near top meaning it was too easy so you can't detect changes or differences
44
Define floor effect
instrument does not register a further decrease in score for low scoring individuals
- scores all near the bottom because it was too hard and can't see real differences
45
Things that help look at statistical significance
- p-value
- precision or confidence intervals
- type 1 and 2 error
- power
46
Things that help look at clinical significance
- how large the difference was
- MCID
- effect size
- sensitivity and specificity
- likelihood ratios
- number needed to treat
47
Explain p-value
- sets threshold for type 1 error
- 0.05 allows for 5% chance the difference is due to error/chance
- no clinical relevance
48
Explain confidence intervals
- range of scores that should contain the population
- expressed as a percentage of how sure you are that the population is between that range
49
Define type I error
- reject the null when it was true
- saying there is a difference when there is NOT
50
Define type II error
- do not reject the null but it is actually false
- saying there is no difference but there is a difference
51
What 3 things affect statistical power
- significance (a)
- effect size
- sample size
52
How does sample size affect power
- larger sample size has greater influence on power
- it is able to detect smaller differences in groups
53
What does effect size do
- determines magnitude of treatment effect
- allows for normalized comparison of results
- accounts for variation
54
Most common effect size variable
- cohens d
55
Small effect size is
0.20
56
moderate effect size is
0.50
57
large effect size is
0.80
58
What would a negative effect size mean
- indicate a decrease
- ie pain goes from 8 to a 2
59
What is gold standard
- zero false positives and false negatives
- what we compare tests to
60
Define sensitivity
- helps determine who has it
- so when we get a negative test we can confidently rule the condition out
- snOUT
61
Define specificity
- helps determine who doesn't have it
- positive test means we can confidently rule IN the condition
- spIN
62
What are the limitations to sensitivity and specificity
- doesn't really tell us a whole lot about the clinical sig
- does not change the probability
63
What is likelihood ratio
- incorporate both sensitivity and specificity
- direct estimate of how much a test result will change the odds of having a condition
64
If a positive likelihood ratio is above 5 this means
- there was a strong or moderate increase in probability
65
If negative likelihood ratio is below 0.2 this means
- there was a strong or moderate decrease in probability
66
What is the pre-test probability
- prevalence rate (what is the chance they have the condition based on prevalence)
- get them from published rates or personal experience
67
Patient history does what
- develops a working hypothesis
68
What is post-test probability
- likelihood that a patient has a specific disorder after considering pretest probability and the likelihood ratio for that specific test used
- must determine if it has crossed treatment threshold
69
What is treatment threshold
- likelihood that the pt has the condition
70
What are the ottawa ankle rules
- reduce the unnecessary use of radiographs in the ER for foot and ankle injuries
EBP 2
flashcards
Decks in class (4)
# Cards
