Market Research MT Class 6

Measurement process

assigning numbers or labels to quantify a concept

Constructs

Brand awareness, patience, risk attitude, implicit attitude, brand loyalty, extroversion

Conceptual definition of contstructs

Theoretical description of what it is

Operational definition of constructs

Translate into something observable; how we measure it

Reliability

degree to which measures are free from random error and provide consistent data; replicability

Validity

are we measuring what we think we are measuring?; broader and more difficult issue

neither reliable nor valid

highly reliable but not valid

highly reliable and valid

Types of reliability

test-retest reliability, split-half/inter-item reliability, inter-rater reliability

Test-retest reliability

same test same people later date

Split-half reliability

give one test, split answers into 2 halves and calculate each participant’s score on each of the halves, measure the correlation

Inter-rater reliability

degree of consistency/agreement between several raters

Types of validity

construct, internal, external, ecological, convergent, discriminant

Construct validity

are you measuring what you think you are

Internal validity

is your measure free of confounds

External validity

does your study generalize to other settings

Ecological validity

does your measure match real life

Convergent validity

does the measure correlate with similar measures

Discriminant validity

is the measure different from similar measures

4 Levels of Numeric Scales

Nominal, ordinal, interval (continuous) and ratio (continuous)

Nominal scale number system

unique labels (0,1,2)

nominal scale marketing phenomena

brands in a category, store types, gender, occupation

Nominal scale sample of appropriate summaries

percentages, frequency, mode

Ordinal scale number system properties

Order of numerals (0<1<2)

Ordinal scale marketing phenomena

ranked preferences, education levels, attitude scale

Ordinal scale sample of appropriate summaries

median, percentile, rank

Interval scale number system properties

equality of differences (2-1=7-6)

Interval scale marketing phenomena

fahrenheit and celsius scale

interval scale sample of appropriate summaries

range, arithmetic, mean, standard deviation

ratio scale number system properties

equality of ratios (2/4=4/8)

ratio scale marketing phenomena

sales (units/dollars), costs, number of customers

ratio scales sample of appropriate summaries

ratios

Dichotomous data (0/1) special case

Nominal, but can be used as nominal, ordinal, interval, or ratio depending on context/purpose

Nominal scales

mutually exclusive, collectively exhaustive (convert to multiple yes/no columns if its a check all that apply scenario)

Nominal scales examples

Where do you live?, What’s your favorite color?, What x do you have? (yes/no, zip codes, brand names)

Ordinal scales def

numbers are assigned to objects based on their order (greater than or less than) or direction

Ordinal scales rule

maintaining the labeling characteristics of nominal scales but add the ability to order data, NO info on HOW MUCH more or less between scale points

Ordinal scales statistical techniques

rank, median, frequency, mode

Interval scales def

distances between numerals correspond to distances between objects

Interval scales rules

differences are meaningful, arbitrary 0 assignment, can’t say 10-degree-day is half as warm as 20-degree-day

Interval scales statistical operation

arithmetic mean, range, standard deviation, median, mode, frequency, rank

Ratio scales def

values assigned possess an absolute zero point so both differences and ratios of scale points make sense

ratio scales example

sales, market share, costs, ages, number of customers; descriptors like “half as much”, “three times as far”, “25 percent more” make sense

ratio scales statistical technique

ratio, arithmetic mean, range, standard deviation, median, mode, frequency, rank

Graphic rating scales

thermometers, uncomfortable to comfortable, smiles

Itemized rating scales

something: unlikely to extremely likely

purchase intent sales

If you did x, would you definitely buy, probably, probably not, definitely not

Likert scales

Strongly disagree to strongly agree

Non-comparative scale

judgement made without reference to another object/concept/person

comparative scale

one object, concept, person compared to another on a scale

Paired comparisons scale

Coke v pepsi, pepsi v sprite, coke v sprite

Rank order scale

order of preference

Semantic differential

2 adjectives, then find the mean response between each set

Constant sum scales

given 100 units, allocate them to which features you want most

Rating scales considerations

nature of construct, type of scale, number of categories, forced vs. non-forcec choice, randomizing order, numbers inside options (phone hours example)

Common scale Response biases

anchoring, extremeness aversion, agreement bias, social desirability bias, random responding