Validity And Reliability Flashcards
What is construct validity?
Does the test measure what it’s supposed to measure?
What is validity definition?
Does the test do what it is supposed to do
-the extent to which a test measures what it is meant to measure
What is reliability
Does the test do what it does every time, is the test and it’s results consistent every time or do they vary?
What is criterion related validity? What is it made up of?
Whether our scale/measure does what it is supposed to do, given our theoretical knowelefge of the construct we are measuring
It is made up of predictive and concurrent validity
What is predictive validity?
Do the scores on the measure predict future behaviour on the criterion variable?
E.g someone with high high school grades would be predicted to achieve high university grades than someone with low scores
What gives us an index or the validity of a predictor test?
The results of the correlation that is run between the first test and the later test
In predictive validity, how do you measure the variance in the criterion that’s caused by the predictor?
We measure using the r2 value
For a good evaluation of predictive and concurrent validities what does r2 and the correlation need to be
Over 16% and over .55
What are some cons of predictive validity?
- time lag (participants may leave)
- redirection of range of variables: only best participants hired
What does concurrent validity compare with?
Indepdent criterion measure
What is concurrent validity?
How well a test correlates with a previous measure that’s been validated. It takes place at the same time, e.g. Students take an iq test and then take a validated one the next day
What is a construct?
A construct is the idea behind an attribute that has been made from research
What is construct validity
Does the test measure what it’s meant to measure? Does a Depression test measure the construct of depression or does it measure the love of kittens?
What is convergent validity?
A subtype of construct validity. It measures to see if a test designed to measure a construct is actually measuring that constructs; it shows that the two predictor variables are related.
What is discriminat validity?
Shows the two measures aren’t related and aren’t measuring what it’s supposed to measure
What is face validity?
Does the test have validity for those completeting it
What is content validity
Does it have items relating to the construct
What three types of validity does a test need to have for a Trinidadian approach to test development?
Predictive, concurrent and construct validity
What are three sources of validity
Factors in the test, individual characteristics and scoring factors are sources of error
What two types of reliability measure the internal consistency of a scale
Split half and inter item reliability
What’s inter rarer reliability
The ratings of two or more people are analysed, the correlations between the two are analysed by two examiners that come to an agreement
What is test retest reliability
Assessing the reliability to see if the test comes to the same conclusion every time
When doing test retest, what measure should be preformed? What value should be a good test retest reliability?
You need to use Pearson r correlation analysis, y
R > .85 is good reliability
What is “>” mean
Greater than
What are some problems with test-retest?
- deciding on an approriate time to retest
- carryover effects
- may be genuine changes in participants over time
- is it practicable to retest participants?
What three conditions must be satisfied for test retest to give an accurate result?
1) first test can’t influence second one
2) reliability measures must be the same for both tests
3) no change in the true score
What are parallel forms of reliability?
A measure of giving different version of the Tesco to the same group of people. The scores of the different versions are then correlated together to see the consistency of the results
How are parallel forms measured?
They are measure using Pearson r correlation
What is internal consistency with reliability?
Do all the items measure the same construct
What is split half reliability
Comparing the performance on half of the test with performance on the other half
What should the reliability be to claim split half realibility
> .75 it should be greater than 0.75
What does cronbachs Alpha measure
Measures the extent to which items responses correlate
What value does cronbachs Alpha have to be to be considthe rthe e internally consistent
More than 0.60
What table in SPSS do you see for the degree to which exch item correlated with the total score?
You look at the item total statistics table. If there is a low score (0.3) it indictatee it’s measuring something different
What should the variance and correlation be for predictive and concurrent validity to be inadequate?
Under 4% and under 0.2
What is content validity?
Does it measure all aspects of a construct?