Attitude Measurement
- Respondents need help to express attitudes / describe images
Itemised Rating Scales
- Researcher first develops a number of dimensions
- Rating scale usually interval scale à evenly spaced points
- Neutral / middle point normally in middle of scale, but there are exceptions
o Example – influence of advertising
- Balanced scales
o Unbalanced scales might be used when it is known that responses will be overwhelmingly in one direction, and need better discrimination in that direction
§ Example à customer service
- Number of points on scale
o 7 is considered optimal by many researchers
o Some researchers suggest 10 (as opposed to 5)
o Issues
§ Distinction possible
§ Ability of respondents to discriminate
§ Degree of discrimination sought
§ Interview medium : telephone (fewer) vs self completion (more)
- Don't Knows
o Distinguishing between genuine mid point response ( neither agree nor disagree) and "Don't Know" can become serious issue
- Odd or even number of points
o Even – eliminate neutral mid point
o Studies: including neutral scale position significantly increases number of neutral responses compared to accepting them spontaneously
o Study: mid point used by respondents trying to reduce effort à exagerate true mid point score
Attitudinal Rating Scales
- Likert scale
o Agree – disagree scale
o 5 point scale
o Full application: sum scores to provide overall attitudinal score for each individual
o Means / std dev can be calc.
o Issues
§ Order effect – order in which response codes are presented
· Bias to left on self-completion scale
§ Acquiescence
· Tendancy for respondents to say yes / agree
§ Central tendancy
· Reluctance of respondents to use exteme positions
§ Pattern answering
· Conflicting answers will identify
- Semantic differential scale
o Bipolar rating scale
o Opposite statements of dimension are placed at two ends of scale and respondents are asked how much they agree.
o Difficult to always determine opposites: modern vs old fashioned / traditional
o Dimensions of similiar meaning should be given with reversed polarity to minimise pattern answering and check internal consistency of responses
- Stapel Scale
o Dimension or descriptor is placed at centre of scale that ranges from -5 to +5.
o Similar to semantic differentials, but not necessary to find accuarte opposite to each dimension
- Numeric scales
o Simple form of scale
o Eg 1 to 10
o End point of scale should be verbally anchored.
- Graphic scales
o Continous bi-polar scale
o Line between two end anchor points
o Visual analoge scale
- Pictorial Scales
o Image of thermometer
- Anchor strength
o Wording of anchor statement is critical
§ Extremely vs very (satisfied)
Comparative Scaling Techniques
- Paired comparisons
o Choose between 2 objects
o Repeat : every object compared with every other object
o Easier than ranking
o Limited to small number of objects
- Constant Sum
o Respondents asked to allocate fixed number of points between set of options to ind relative importance etc
o Reuires some mental agility on part of respondent.
- Card sorting
o Sort cards into groups,
o Then rank each group
- Q sort
No comments:
Post a Comment