MKTG 465 04 - measurement scaling

Measurement scaling

“rules for assigning numbers to objects to represent quantities and qualities of attributes” (Bennett 1995)

interval

• distance between 2 numbers on a scale. refers to whether difference between values is equal and meaningful.

  • 0C, 10C, 20C = intervals are equal + meaningful

categorical data - words

nominal and ordinal

scale data - numbers

interval and ratio

nominal

“namesake” - numbers assigned to objects are classifying them, but have no numeric meaning

nominal example

You are:

_______ Married (1)

_______ Not married (2)

• the numbers 1 and 2 have no significance other than to categorize married and not married.

Ordinal

• nominal scales that can order data

• strictly indicators of rank order. numbers do not imply absolute quantities, nor do they imply equal intervals

ordinal example

ranking drinks from most preferred the least preferred.

a. Coca Cola

b. Pepsi

c. Sprite

d. Mountain Dew

e. Seven Up

• numbers have meaning, intervals not equal

interval

• ordinal scales with equal intervals between points to show relative change

• intervals have meaning. absolute scale values do not.

interval example

lecture satisfaction on a scale of 1 - 5, 1 being very dissatisfied and 5 being very satisfied. intervals are equal and meaningful.

ratio

• interval scales with a meaningful zero point so that magnitudes can be compared

• zero value indicates absence of a construct

• a number with real value = age, percentage

ratio example

• describe your age in years → number is absolute and meaningful

variables

• Concepts derived from the grouping of salient attributes

• Concrete and observable

  • Gender, Income, Age

constructs

• Mental conceptions derived by mutual agreement for a specific purpose

• Abstract and unobservable

  • Personality, Brand loyalty, Lecture satisfaction

attitude trilogy

• attitude = what you think (cognitive) + how you feel (affective) + what you do (conative)

measuring marketing constructs (churchill)

1. specify construct domain

2. generate item statements

3. establish measurement criteria

4. assess accuracy + stability of construct measures

1. specify construct domain

• define central idea under study, define boundaries

• source: academic literature

2. generate item statements

• describing construct

• source: academic literature + subject matter experts

3. establish measurement criteria

• assigning value to construct

• EX: using a 7 point equal interval rating scale

• ranking: rank order objects, ordinal scale, shows order, not magnitude

• ratings: descriptive and numerical

3.1: descriptive rating

words only, ordinal

3.2: numerical rating

numerical (scale), interval

4. assess accuracy and stability of construct measures

→ MO = MT + MS + MR

MO = Observed measure

MT = True score

MS = Systematic/Constant error

MR = Random error

• extent to which MO is accurate and stable:

  • measurement validity

  • measurement reliability

measurement validity

Correctness / Measuring what was intended to be measured

• The extent to which MO = MT

• Can only infer the validity of a measure

types of measurement validity

face (weakest)

content (Middle)

construct - convergent and discriminant (strongest)

face validity

“looks right”, researcher judgement, no guarentee

content validity

supported by literature, previous studies

construct validity

• empirical evidence matches theory

• convergent validity: measures of construct correlate

• discriminant validity: measures of different construct DO NOT correlate

measurement reliability

• consistency or stability, extent to which MR = 0

• minimize random error

• 2 dimensions: repeatability and internal consistency

repeatability

Replication to determine similarity between two separate administrations of the scale

• test-retest

• equivalent sample

repeatability: test-retest

same instrument, same sample, diff time.

• similar results = reliable

repeatability: equivalent sample

diff sample, similar results = reliable

internal consistency

Consistency within/between groups of questions comprising the scale

internal consistency: equivalent form

2 similar instruments = similar results

internal consistency: split half technique

divide items into 2 groups, correlate the halves

• high correlation = high internal consistency

internal consistency: coefficient alpha

• cronback 1951

• measures internal consistency

• alpha >= 0.6 → acceptable reliability

• higher = better

low alpha

too much random variation