Week 2 Flashcards
Measurement
Assigning numbers or labels to things to describe them.
Example: Measuring how tall someone is (like saying they are 170 cm tall).
Scaling
Creating a range of values to show different levels or degrees of something.
Example: Using a scale of 1 to 5 to show how happy someone is, where 1 means “not happy at all” and 5 means “very happy.”
Comparative Scaling
EVALUATING or RANKING 2 (or more) items BASED on their relative DIFFERENCES DIRECTLY.
(instead of seperate evaluations of each one)
Comparative Scaling - 1. Paired Comparison Scaling
Respondent needs to select one item (out of 2) according to some criterion
* Heineken or Grolsch * Heineken or Corona
Comparative Scaling - 2. Rank Order
Respondents need to order several objects according to some criterion
- Grolsch, 2. Heineken, 3. Corona
Comparative Scaling - 3. Constant Sum Scaling
Respondents allocate a fixed amount of points among several items according to some criterion
* Grolsch 50, Heineken 25, Corona 25
Each stimulus object is scaled independently of the other object.
Non comparative Scaling
Evaluating ONE item SEPARATELY , without direct comparison to other items.
Non comparative Scaling - 1. Likert Scale
Respondents need to indicate level of agreement to a series of statements
How much do you like taste of Heineken 1-5 (very to not at all)
Non comparative Scaling - 2. Semantic Differential Scale
Rating scale with endpoints associated with bipolar labels
unpleasant ……..:……:…..:……:……:………pleasant
Non comparative Scaling - 3. Slider Scale (continous rating scale)
Respondents rate the items by placing a mark at the appropriate position on a continous scale 0 - 100
Multiple item measurement theory
Using several related questions instead of just one.
To increase accuracy -> reliability
MODEL:
Construct (big picture (e.g. Buying tendancy)) = Related questions scores / Number of related questions
Xo = Xt + Xs + Xe
*Xo = observed score (persons score on BT)
*Xt = true score (real BT)
*Xs = systematic error (non random error, social bias)
*Xe = random error (being distracted, mood)
Reliability (consistency)
Same results when the measurement is repeated (under the same conditions)
Example: If you use a weight scale multiple times and it shows the same weight each time you step on it, the scale is considered reliable
Validity = Accuracy
Actually measures what it is supposed to measure.
Example: If a test claims to measure intelligence, but it primarily assesses athletic skills, it lacks validity. A valid intelligence test should evaluate various cognitive abilities.
Reliability = Validity ?
A measurement can be reliable but not valid.
For instance, a bathroom scale that consistently shows the wrong weight is reliable but not valid.
2 Methods of assessing scale RELIABILITY
- Test/Re-test (stability)
- Internal Consistency (Cronbach’s α)
- way to check if all the questions are measuring the same thing
Cronbach Alpha THRESHOLD = closer to 1 the better (0.70 acceptable, 0.90 practical application)
*Improve α by deleting items with low correlation
3 methods of Assesing VALIDITY
- Content
-Measures if a test includes all of the relevant topics.
* Example: A math test should cover all areas of math, not just a few. - Criterion
-Assesses how well one test predicts outcomes of another test.
* Example: SAT scores predicting college success. - Construct
-Checks if a test truly measures the concept it claims to measure.
* Example: A self-esteem test should measure self-esteem, not something else like anxiety.
Construct Validity: 3 types
- Convergent Validity
- whether two or more tests designed to measure the same concept yield similar results.
*Example: Two self-esteem tests should give similar results. - Discriminant Validity
- Checks if a test does not measure something it shouldn’t
*Example: A self-esteem test should not give similar results to an intelligence test. - Nomological Validity
- Looks at how a test behaves in line with theory
*Example: A test of self-esteem should relate to extrovertedness if the theory says they are connected.
Factor Analysis (definition + goal)
Definition: Reduces complexity by grouping related variables into a few core factors
Goal: Tries to capture as much info from the OG variables as possible (in the new few factors) with as few factors as possible
How? Group together variables that are highly correlated.
Variables in the same group = high correlation
Variables in different groups = low correlation
Higher loading = higher correlation
2 Core Objectives of Factor Analysis
- Data Reduction
- To simplify a large set of variables by reducing them to a smaller number of underlying factors or dimensions. - Identify Patterns
- To find and group related variables that measure the same underlying concept.
Whats Barletts test of spehercity and KMO
Barletts test of spehercity = tells us if at least some of the variables are correlated with eachother (p<0.05)
KMO - it indicates how intercorrelated the variables are = the higher MSA the more correlated they are (>0.50)
What is Eigen Value
Amount of variance explained by factor
(how much info of variables explained by each factor)
Sum of all eigenvalues = number of variables
Eigenvalues above 1 = give the most information/variance
Rule for selecting factors: Eigenvalue > 1
We want each factor to explain the variance of at least a single variable (Note: Each variable has a variance of 1)
Communalities Extracted (threshold + what happens if low)
How much of the variance of each variable is captured by the extracted factors. (from the 3 factors we identified)
Communalities need to be = or > than 0.30 to not be removed
If communality is very low (say < .30), the item is “quite unique” since it correlates weakly with other variables. Such variables should be removed, as it is definitely measuring “something else.
Interpretation of the Rotated Solution
Presents correlation between factor and each variable.
absolute values !!
Look for simple structure: each variable (hopefully) loads high on 1 factor and low on other factors
They need to be above 0.5 and grouped together with other variables within a factor (otherwise we exclude) !!!
Steps of Factor Analysis
- Is Factor Analysis appropriate?
a. Bartletts test of spherecity
b. KMO- How many factors?
a. Total Variance Explained (eigenvalues)
b. Scree Plot (elbow rule) - Seeing if every variable is correlated (uniqueness)
a. Communalities - Grouping variables into factors
a. Rotated Component Matrix - After a factor analysis, calculate Cronbach’s alpha for each identified factor to assess internal consistency (reliability
a. Cronbach’s alpha if item deleted
B) Factor Scores - How many factors?
Perceptual Maps
Help visualize brand positioning and comparisons in a simple way.
- Crucial for making informed marketing decisions.
*A joint-space (or product-market) map
Displays both (brand) perceptions and consumer preferences on a single map
- A very useful tool for understanding current position relative to competitors and to re-position the firm’s offerings.
Variance
How dispersed the data points are with respect to the mean
(or amount of info)
Preference Model
- Consumers are represented as “ideal points” in multidimensional space.
- Brands are preferred when they are closest to the consumer’s ideal point on the perceptual map.
