Statistics Flashcards
Hierarchy of evidence
- SR + MA
- RCT
- Cohort
- Case-control
- Cross sectional survey
- Case report
- Expert opinion
- Anecdotal
Cohort studies
Compare groups prospectively
Case-control studies
Compare groups retrospectively
Cross-sectional surveys
Look at a group at a specific point in time
**Typically utilized in developmental psyc
Nominal
- Labels that are mutually exclusive
* *Male + female
Ordinal
- Rank ordering, distance between ratings not equal
* *1st, 2nd, and 3rd place
Interval
- Equal intervals between ratings, no true zero point
* *Temperature
Ratio
- Equal intervals, has true zero poin
* *10 meter walk time
Reliability
Consistency, dependability
Random errors limit…
Reliability
Systematic errors limit…
Validity
SEM
Repeated measures on the same instrument tend to be distributed around the “true” score
Large SEM =
Small SEM =
Low reliability (large) High reliability (small)
Intra-rated reliability
Consistency within the same rater
Inter-rater reliability
Consistency between raters
Validity
Extent to which a test measures what it is purported to measure
Construct validity
How well the test measures The abstract construct it is supposed to measure (pain, intelligence, QOL)
Content validity
How well the content of the test matches a content domain associated with a construct
**Determined by expert consensus
Face validity
Instrument appears to test what it is supposed to
Criterion-related validity
Compares the test with other measures already validated
**Compared to gold standard
Concurrent validity
Comparison between the test and another measure administered at the same time
Predictive validity
Comparison between the test and another measure administered in the future
Floor effect
A measure’s lowest score is unable to assess a patient’s level of ability
Ceiling effect
A measures highest score is unable to assess a patient’s level of ability
Normative data
Represents scores pulled from the literature to provide “normal” values within a population
Cut-off scores
Designates a positive or negative test outcome
MDC
Minimum amount of change in a patient’s score that ensures that change is not the result of measurement error
Minimal clinically important difference
The smallest amount of change in an outcome that might be considered important by the patient or clinician
Sensitivity
- True positive rate
- Likelihood that someone with the condition will be positive on the diagnostic test
Specificity
- True negative rate
- Likelihood that someone who doesn’t have the condition will negative on the diagnostic test
Positive predictive value
Percentage of people who are positive on the diagnostic test who have the condition
Negative predictive value
Percentage of people who are negative on the diagnostic test that do not have the condition
Positive likelihood ratio
Indicates how many times more or less likely a positive result will occur in someone with the condition that in someone without the condition
- *How much the odds of a disease increase when a test is positive
- *Increase the probability of having the disease associated with having the finding vs not
+LR values
> 10 Large
5-10 Moderate
2-5 Small
1 Neutral
Negative likelihood ratio
Indicates how many times more or less likely a negative test restful will occur in someone without the condition than in someone with the condition
- *How much the odds of having the disease decrease wit ha negative result
- *Decrease the probability of having the disease associated with having the finding vs not
-LR values
1 Neutral
.2-.5 Small
.1-.2 Moderate
Descriptive stats
- Summarizes a sample rather than a population
* *Does not allow us to make conclusions of data from the sample
Nominal - measures of central tendency
Mode
Ordinal - measures of central tendency
Median
Interval/ratio (not skewed) - measures of central tendency
Mean
Interval/ratio (skewed) - measures of central tendency
Median
Range
Difference between the highest and lowest values
**Does not give insight into distribution of scores
Percentiles/quartile
A value below which a given percentage of scores fall
Standard deviation
Measures of the distribution of scores around a mean
Low and high standard deviation
Low - tend to be close to the mean
High - tend to be spread out over a wider range of values
Intraclass correlation coefficient
- Measures of reliability of ratings
- Rangers rom 0 to 1 (low to high agreement)
- Generally, ICC > .75 is good reliability
Inferential statistics
Techniques that allow us to use samples to make generalizations about the populations the samples represent
Null hypothesis
Assumes there is no relationship between variables
Experimental hypothesis
Assumes there is a relationship between the variables
Type 1 Error
Alpha
False +
Claiming there is a difference in groups when in fact there is no difference
Type II Error
Beta
False -
Claiming there is no difference/failing to reject the null when there is a difference
Power
Strength of the treatment effect
1-B
P-value - level of significance
Chance that the results occurred by random chance
P-value < alpha
Null is rejected
**Lower the p-value, the less likelihood that the results occurred by chance
P-value > alpha
Null is not rejected
**Higher the p-value, the more likelihood that the results occurred by chance
Parametric
- Representative of the population
- Follows normal distribution
- *INTERVAL + RATIO
Non parametric
- Not representative of population
- No normal distribution or randomization
- *NOMINAL + ORDINAL
Confidence interval
How confident you are that the value falls within that interval
**Wider the interval, the higher the confidence
Coefficient of determination (R2)
Describes the amount of variance in the dependent variable that can be explained by the variance in the independent variable
2 related scores
>3 related scores
2 - P - paired test, NP - Wilcoxon signed ranking
3 - P - ANOVA, NP - Friedman
2 ind groups
>3 ind groups
2 - P - unpaired ttest, NP - Mann-Whitney U test
3 - P - ANOVA, NP - Kruskat Wallis
