Reliability (what is it)
reliable=doesn’t fluctuate
- true score: the ‘real value’ on a given variable
- measurement error: any deviation from the true score
Assessing reliability
- reliability is essential
- we can calculate the person product- moment correlation coefficient on the results of the measure taken from the same people several times
types of reliability
test
- retest reliability= the same person gets similar scores on repeated tests
- internal consistency reliability= people’s scores on individual tests are similar to each others
- iterrator reliability=comparing 2 observer’s ratings of behavior
- reliability is not the same as accuracy
Construct validity
- my measure ‘true?!’ How well does my operational def and measurement match reality? Does it actually measure what it is intended to measure?
Indicators of Construct Validity
-
-
-
-
-
-
-
-
-
-
- content validity
- concurrent validity
- convergent validity
- predictive validity
- discrimination validity
- Face validity
Content validity
Content validity evaluates how well an instrument (like a test) covers all relevant parts of the construct it aims to measure.
Predictive validity
Predictive validity is the degree to which test scores accurately predict scores on a criterion measure.
Concurrent validity
does it predict some current behavior
convergent validity
scores on this measure are related to scores on other measures of the same/similar constructs
discriminant validity
scores on this measure are not related to scores on a measure of some other construct
Reactivity of measures
- reactivity: when measuring a variable changes the thing you’re trying to measure
Measurement scales
- normal scales: categorical; no numerical or quantitative properties
- ordinal scales: rank order levels of variable from first to last
- interval scales: 0 does not mean nothing, - exists
- ratio scales: zero means nothing, can’t be -
Face validity
does it look like what it’s supposed to look like?
content validity
Content validity evaluates how well an instrument (like a test) covers all relevant parts of the construct it aims to measure.
Predictive validity
Does it predict some future behavior?
concurrent valididty
the degree in which the scores on a measurement are related to other scores on other measurements that have already been established as valid.
convergent validity
scores on this measure are related to score measures of the same/similar constructs
Cronbach’s alpha
calculate the average of all possible split half reliability coefficients
item-total correlations
see how well each specific item correlates with the total score. Items that don’t correlate well with the total score can be eliminated
