Interpreting Mesaurements Flashcards
Validity
Did the test measure what it says it measures?
Reliability
How confident can I be about the measurement
A value that quantifies consistency of a tool
Necessary, but not sufficient, for a measure to be valid
Measurement scales
Categorical scales:
Nominal (names)
Ordinal (has order but not consistent intervals)
Continuous scales
Interval (has order and equal intervals)
Ratio (has order, interval, true 0)
Nominal and Ordinal Scales
Nominal scales:
Categorical
No order
Ordinal scales:
Categorical
Are ordered
Continuous Scales
Interval scales:
Are continuous
Have equal intervals
Allow for mathematical operations
Ratio scales:
Are continuous
Have equal intervals
Zero means absence
Reporting Options by Scale: Nominal and Ordinal
Reporting for a group (“descriptive statistics”)
Nominal: Frequency, tallies, counts, percentage, and mode
# in each group, % of total in each group, most frequent, pie chart
Ordinal: (All of the above) + Central tendencies (mode, median) and Range (including interquartile range) , box and whiskers plot,
Reporting Options by Scale: Interval and Ratio
Reporting for a group (“descriptive statistics”)
Interval or ratio: Range (including interquartile range), Central tendencies (mode, median, mean), Variability (Standard Deviation), Box and whiskers plot
Reliability measurements
Test–retest reliability
Inter-rater reliability: Consistency between two different raters
Intra-rater reliability: Consistency between same rater
ALL measurements contain error!
Avoid saying a test is reliable or unreliable use coefficient instead
Interclass correlation coefficient (ICC)
Used for continuous data
(Variability between subjects – Variability within subjects) / (Variability between subjects)
1 is perfect reliability
0.9 is excellent for clinical measures
0.75–0.9 is considered good
0.5–0.75 is considered moderate
Doesn’t give the error expected around a specific measurement
Doesn’t give a way to interpret how much change needs to occur to be beyond error of the tool
SEM to CiM
Standard error of measure (SEM): estimate of the average variability expected around a measurement
Use SEM to determine 95%CI around a
measurement (confidence in measure, or CiM)
SEM and CiM quantifies the potential error
(variability) around a measurement taken - used to calculate MDC
MDC
Minimum amount of change required to
exceed measurement error
MDC takes into consideration the error associated with two measurements (based on ICC and SD for a sample)
Determines what change would be beyond the error associated with taking two measurements.
MCID
Minimal clinically important difference
Smallest change that would be important to the
patient
Relationship between measurement and function
Established for a patient group/population
