Key concepts - measurement Flashcards
To understand key concepts of meaurement (21 cards)
What are ways to improve reliability?
1) Standardize measurement methods
2) Train and certify interviewers, clinicians
3) Refine instruments (improve measurement methods/ pilot tests)
4) Automate instruments (remove observer variability)
5) Repeat measurements in different populations (provides mean value, reduces variability)
What is reliability?
Repeatability of results based on the degree that they are free from random error
What is precision?
Precision is the level of detail or exactness in a measurement
How is precision related of reliability?
More precise measurement = less random error and less variation in repeat results
How do we assess reliability?
1) Test-retest
2) Inter-rater or observer reliability (Kappa, % agreement, .60+ is acceptable)
3) Internal consistency - are items measuring the same concept - Cronbach alpha (.70 is acceptable)
What are sources of variation in tests?
1) Biological variation (eg, age-related differences in lipid levels and BP)
2) Analytical variation (technical factors affecting the reproducibility of findings)
3) Clinical or disease-related variation (eg, presence/absence and extent of disease)
What is validity?
How close a measure is to the true state of nature
A valid meaure has low to no bias / systematic errors
For a diagnostic test, what is validity?
The ability of a test to indicate which individuals have disease and which do not
Assessed with sensitivity, specificity, PPV, and NPV
What is sensitivity?
The probability of a positive test given that the person truly has disease
as determined by a gold standard test
What is the equation for sensitivity
Probabilty of a true positive (TP) = TP/(TP+FN)
FN = false negative (these are actually positive too). So basically the
In a 2x2 table with test on left and disease on right a/(a+c)
Are sensitivity and specificity affected by disease prevalence?
They are the **fixed ** characteristics of the test, not the disease.
Are predicitive values affected by disease prevelance or test specificity/sensitivity?
Yes
The higher the sensitivity, specificity, and disease Incidence, the high
What is positive predictive value?
The proportion of people with a positive test who truly have disease
TP/(TP + FP)
In a 2x2 table (test on left; disease on top): a/ (a +b)
Negative Predictive Value
The proportion of people with a negative test who truly do not have disease
TN / (FN + TN)
2x2 table (test on left; disease on top: d / (c + d)
What is an outlier
Observations far away from the center of the data by some standard
Per Tukey, any point above upper fence or below lower fence of IQR
if IQR = 75th-25th then upper fence = 75th percentile + 1.5IIQR and lower is 25th - 1.5IQR
An outlier of concern is…
both high influence and high leverage
What is leverage?
A measure of how unusual a point is in terms of the predictors (having an unusual combination of predictor values)
What is Influence?
A point that, if removed from the model, would substantially change the fit of the model or the coefficient estimates.
Assessed with Cook’s distance
(Cook’s distance is an observation’s standardized residual and its leverage) Values >1 indicates a point has substantial influence. Values <.5 are unlikely to influence the model.
What is Power?
The probability of detecting a true effect if one exists
What is Beta (β)?
Probability of not rejecting a false null hypothesis.
Usually set at .8 in a power calculation
