RELIABILITY AND VALIDITY OF MEASUREMENTS Definitions Reliabilityconsistent reproducible

RELIABILITY AND VALIDITY OF MEASUREMENTS

Definitions Reliabilityconsistent, reproducible, dependable Validitymeasures what it says it measures

Reliability Measurement Error Reliability Coefficients Types of Reliability

Measurement Error Observed score = true score + error Measurement Error = true score – observed score Reliability estimates measurement error

Sources of Measurement Error Systematic- consistently wrong in the same amount Random- chance rater measuring instrument variability in what you are measuring

Reliability Coefficients True score variance + error variance as error increases, coefficient increases coefficient ranges from. 00 to 1. 00 <. 50 poor. 50 to. 75 moderate >. 75 good

Types of Reliability: Test-Retest Get the same results every time you use the test Intervals between testing- long enough to avoid fatigue or remembering the answers but not so long that natural maturation occurs Intraclass correlation coefficient (ICC)

Another way to Test- Retest Alternate forms: different versions covering the same content (SAT, GRE) correlation coefficient is used

Types of Reliability: Internal Consistency Are all questions measuring the same thing? Split-half: correlation of two halves of same test (odds and evens) Spearman-Brown prophecy Cronbach’s alpha: essentially an average of the all the possible split-half reliabilities, can be used on multiple choice. When used on dichotomous scores, called Kuder-Richardson 20

Types of Reliability: Raters Intra – rater: stability of one rater across trials Use ICC for both types Inter-rater: consistency between two raters

Validity vs. Reliability Types of Validity

Validity vs. Reliability

Generalizability External validity- the test is valid if used with the intended population The test is reliable if used in appropriate context and as directed for its given purpose

Face Validity appears to test what it is supposed to measure weakest form ok for ROM, length, observation of ADLs

Content Validity covers the entire range of the variable and reflects the relative importance of each part based on expert opinion, needs to free of cultural bias Test of function- 20 questions on brushing your teeth, 1 question each on mobility, bathing, dressing VAS vs. Mc. Gill Pain Questionnaire

Criterion-related Validity target test compared to gold standard concurrent target test is taken at the same time as another test with established validity predictive examines whether the target test can predict a criterion variable

Construct Validity ability of a test to measure a construct based on a theoretical framework what would you include for a test on “wellness”?

Ways to establish construct validity Known groups Convergent comparison Divergent comparison Factor analysis

Remember The reliability and validity of a test measurement is not the same thing as reliability and validity of a research design.