More Validity And some reliability. More validity Construct validity Content validity Face validity Concurrent validity Predictive validity Discriminant.

1 More Validity And some reliability ...
Author: Adela Ford
0 downloads 4 Views

1 More Validity And some reliability

2 More validity Construct validity Content validity Face validity Concurrent validity Predictive validity Discriminant validity Convergent validity

3 Construct validity gone awry Homosexuality –Classified as a mental disorder by APA until 1973 –Freud believed it was an immature stage in sexual development (not an illness) Due to overidentification with the mother –Other psychoanalysts argued it was an illness due to overinvolved mother and rejecting father behavior

4 More validity Content validity –Does the test or measure have the content that represents the construct –Does it have content that doesn’t represent the construct Content validity of a measure of autism? –What should it include?

5 Concurrent validity Is the measure associated with other related qualities or events that would occur simultaneously Concurrent validity of the SAT? –What would you expect it to be related to?

6 Predictive Validity Does the measure predict, at a later date, to something that should be related? Your SAT scores and…? What are your college grades a measure of? What sort of predictive validity will they have?

7 Convergent Validity Is the measure correlated with other similar measures? If it’s a measure of racism, will it be correlated with other measures of prejudice and discrimination?

8 Discriminant Validity Does the measure NOT correlate with things it should NOT correlate with? Not correlate is not the same as negatively correlated Can be hard to establish E.g. a measure of math skills should be uncorrelated with a measure of verbal skills –If we believe these are unrelated constructs

9 Face Validity Does it seem to measure what it is measuring This can be good or bad –Good if it helps people zero in on the questions –Bad if measuring something people don’t want to be honest about-like racist attitudes Good face validity does not mean a measure is good

10 In Class exercise Come up with a research idea for the topic area you have been given—anything you want!! Identify two variables or more Say how you might measure them

11 Reliability Is a measure or instrument consistent? –Across time –Across observers/raters –Across questions Without reliability there is no validity

12 Test-Retest Reliability Is it consistent over time? Does NOT mean scores should be “identical” over time Means relative rank order of each person in a group should be similar across time points

13 Test-retest reliability E.g. The Hare Psychopathy checklist –A group of people takes it twice –We see if their scores are related over time The Beck depression inventory –A group of people takes it twice –We see if their scores are related over time How long a time? –Depends on the construct

14 Test-retest reliability Measured with the correlation coefficient R R is a measure of the relationship between two continuous (or rank ordered) quantitative variables –E.g. Height and weight

15 The correlation coefficient R A number between -1 and +1 Closer to 1 or -1 is stronger relationship Positive correlation: if one variable has high scores, the other also tends to have high scores –And if it has low scores, the other also does Negative correlation: when one has high scores, the other has low scores

16 Positive Correlation

17 Negative Correlation

18 No correlation

19 Test –retest reliability Is calculated using the correlation coefficient R If someone has high scores on a depression inventory one week, we expect the same thing again two weeks later –Relative to other people who took it –If the measure has reliability over time

20 Test-retest reliability Key for most surveys of mental illness, personality traits, intelligence, etc. We expect some stability in these qualities If we don’t expect stability, test retest reliability not necessary or useful But if we expect stability and don’t have it –Then our measure is bad How much stability? –Usually at least.70 or higher, depending on time length