CCJS300 Final Scales & Indexes

Scale

data reduction technique; goal is to end up with single numerical value

Steps in scale selection

item selection

bivariate relationships

multivariate relationships

scale scoring

handling missing data

Item selection

developing an initial pool of potential items

**face-logical validity**: researcher’s opinion on what measures the topic

Bivariate relationships

sees if pools are significantly related to each other

remove items that aren’t related

**correlation**: measure of association

Multivariate relationships

see if each variable adds something unique to the scale

eliminate items that are perfectly/almost perfectly correlated to each other

Scale scoring

**summated scale**: assign points for every answer & add the points up; range of scale depends on items

variables weighted the same unless one deemed more important

**negative items**: items worded in the opposite to avoid acquiescent response set

Handling missing data

exclude all cases of missing data if feasible

examine missing data for a pattern

assign mean/median score to missing data

use “undecided” for missing data

Likert scale

summated scale; assigning scores for each answer & adding them up

Guttman scale

order of items makes a difference; allows prediction of previous answers

dichotomous response categories

Coefficient of reproducibility

assesses consistency

1-(# of errors/# of responses) > 90%+

Bogardus social distance scale

type of Guttman scale

indicates respondents’ tolerance of different demographics

Q-sort

subject presented with several hundred statements/cards

rules of how many cards go into a particular pile; measures of internal consistency

goal is to create different scores

Semantic differential scale

subject presented with opposite choices on a continuum

easier version of Q-sort

Factor analysis

statistical tools that help researchers develop scales

identifies unique sources of variation among variables

each column presented = a factor

numbers = factor loading (like correlation coefficient but no directoin)

Scaling

attempts to increase the complexity of the level of measurement of variables from nominal to at least ordinal

Scale vs. index

scale = concerned with attribute of a concept

index = involves many dimensions/scales

Arbitrary scales

developed by the researcher based primarily on face validity & personal judgment

Crime seriousness scales

procedures that assign weight or severity ratings to various crimes

came as result of UCR criticisms

Sellin-Wolfgang index

procedure for assigning crime seriousness weights

accounts for both quality & quantity of crime

Types of seriousness scales

simple rating scales

magnitude scales

Simple rating scales

respondents rate crime on a scale of 1-9; average used as measure

Magnitude scales

measure public rankings of the degree of relative seriousness of various crimes

Prediction scales

attempts to forecast crime commission or success/failure on probation/parole

social prediction table

statistical prediction table

Greenwood’s Rand Seven-Factor Index

self-report scale of inmates that resulted in an arbitrary scale to predict recidivism

aimed at selective incapacitation

Greenwood’s 7 variables

prior conviction for same charge

incarcerated more than 50% of preceding 2 years

convicted before age 16

served time in a state juvenile facility

drug use in preceding 2 years

drug use as a juvenile

employed less than 50% preceding 2 years

Advantages of scales

enables more exact measurements

easier replication & longitudinal measurement

Disadvantages of scales

may not actually measure what it’s intended to measure, but a person’s attitude/behavior