chapter 4 - scales of measurement

reliability

the extent to which a measure produces stable, reliable, consistent scores, direct function of the amount of measurement error present

test-retest reliability

assessing reliability by administering a measure to Ps at 2 different times and correlating the results, high consistency=high reliability

internal consistency

degree to which the total set of items in a multiple-item measure behave in the same way

cronbach’s alpha

average inner-item correlation for all items in questionnaire

validity

the extent to which an assesment measures what it is intended to measure, measures lacking validity are untrustworthy

construct validity

extent to which a measure accurately represents the construct it’s said to measure, tied intimately to operational definitions and can’t be established or destroyed with a single study (only multiple studies)

content validity

the extent to which a measure covers a representative sample of the domain of B to be measured, like measuring height from the waist up

criterion validity

a measure related to some behavioral outcome or criterion already established, assesses whether a measure:

1. can accurately predict future B (predictive validity)

2. meaningfully relates to a presently established measure of the same construct, ie. narcissism test

convergent validity

extent to which scores on an assessment relate to scores on a different assessment presumed to measure a theoretically related construct, ie. dominance vs. leadership vs. shyness

discriminant validity

extent to which scores on a measure do not relate to scores on assessments presumed to measure theoretically different constructs, ie. relationship with a sociability scale like number of friends and neuroticism

face validity

addresses whether or not a test looks valid on the surface, not validity in the technical sense but refers to what the test appears to measure, important for engagement, perceived trustworthiness of measure, and deception studies

nominal scale

used for classification/labeling without any order, any number used is only as an identifier, ie. political affiliation 1=D 2=R 3=I

ordinal scale

data is ranked/ordered but differences between ranks are not equal, ie. class ranking or college basketball top 25

interval scale

equal intervals between values but no true zero, so having a score of zero doesn’t mean nothing happens there, ie. temperature or iq scores

ratio scale

equal intervals between values but with true zero that is meaningful, ie. time or income

discrete (categorical) variable

variables that take on a finite number of values and are mutually exclusive (if you are one, you can’t be the other), results in count data/whole numbers, ie. average children per household is 2.3 which means 2-3 kids not literally 2.3 kids

continuous variable

variables that take on any value including decimals, nominal variables cannot be continuous, ie. weight or time

X,Y, Z

uppercase english letters represent variables

X_1, Y₁

uppercase letter with subscript represents individual values of a variable

sigma (Σ)

symbol for summation which means “sum what follows”