Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
24 Cards in this Set
- Front
- Back
Reliability |
- consistency - reliability coefficient normally assumes values of 0.00-1.00 |
|
What types of things are tested for reliability? |
1. observers or raters 2. tests over time 3. different versions of the same test 4. a test at one point in time. |
|
internal consistency |
- inter-correlations among individual items of a knowledge test or trials on the same day. |
|
what is a sufficient number for reliability? |
.70 or higher |
|
Types of Internal consistency statistics? |
1. Split Halves - compare odd to even 2. KR-20 - better than split halves. It randomnizes the order of items (uses to measure correct vs incorrect response). 3. Cronbach's alpha - same as KR-20 but can be used with scaled items like Likert scale of strongly agree to strongly disagree. |
|
Stability (test-retest) |
- results would be the same if administered to the same person again at a later time. - coefficient generally decreases as the time between testing increases - |
|
Parallel-forms |
- administer two forms of the same instrument to a single group within a short time period between testing - correlate the results |
|
Parallel-forms reliability measures what? |
- stability across forms |
|
Inter-Rater or Inter-Observer Reliability |
-measure if different observers are consistent -establishes percent of agreement -can use correlation (with continuous ratings) |
|
Inter-rater reliability statistics |
1. Kendall coefficient of concordance - each ranks the thing being evaluated. The higher the agreement between raters the closer to +1 2. Cohen's Kappa - used for nominal level data (can be used with more than 2 raters, but no in SPSS) 3. Intraclass correlation coefficients (ICC) - for interval or ratio level data. |
|
ICC |
- based on repeated measure ANOVA - 6 different formulas for different situations - can do more than 2 repeated measures (correlation only two) - determines absolute and relative reliability |
|
how do you assess observer reliability? |
1. unit by unit agreement requires two or more observers to agree on individual instances of response being measured 2. # of units on which both agree is counted A 3. # of units on which both agree is counted D 4. index = (A/A+D) x 100 |
|
list observer issues |
1. bias - may see what they want to see or their own preconceived idea 2. observer should report data not inferences 3. use of trained observers |
|
how do you measure reliability with Pearson r |
- add paired t-test to determine the magnitude of the differences - regression - absolute reliability established if slope of the line is close to 1 and intercept is close to 0. - calculate SEM for paired data |
|
Standard Error of the Measurement (SEM) |
- can be used to estimate the range a score would fall if a given measured object was re-measured (confidence bands) - if reliability is high, SEM will be small. |
|
can an instrument be reliable without being valid? |
yes
if the instrument is valid it will be reliable |
|
what is validity |
a test or survey is valid when it measures what it is supposed to measure |
|
how does one gain content validity? |
- literature or expert panel - subjective measurement |
|
2 types of criterion based validity |
1. establish criteria called gold standard to compare with new measure 2. compares between two measures |
|
concurrent validity |
- two measure at the same time (exam scores and overall GPA) - measured by a correlation coefficient (r) |
|
Predictive validity |
- measurement correlated with a future measurement (GRE scores used to predict graduate school success) |
|
construct validity |
-psychological constructs measured such as self-esteem -phsical measures such as the broad construct of strenght |
|
steps to a factoral analysis |
1. group of variables analyzed for inter-relationships (correlation matrix) 2. underlying factors determined by reviewing variability of values in data set that can be accounted for by the factor). 3. factors are extracted-removal of factors that don't measure up 4. rotate factors-maximize the appearance of differences between factors to create factor loadings 5. examine rotated factors to determine simple structure. Retain variables in each factor that loaded above a cutoff point (commonly .30) 6. name and interpret factors based upon the remaining variables in each factor |
|
reliability and validity are what? |
just estimates. not absolute values |