Home / Summaries / Marketing Models / scale-topic-capture

Measuring customer satisfaction and factor analysis - Measurement theory

Q: How can you cancel out the measurement errors in surveys?

By asking multiple questions on the same topic so that the scale reflects the basic nature of the concept.

Q: What type of correlations do you expect when the survey is good?

High correlations between the grouped questions about one topic.

Q: What index is used to capture reliability and how is it calculated?

Coefficient alpha . Ranges from 0-1 and best when above 0.7 Standardized : With p = number of items in the scale (i.e. how many questions in the group) With the denominator = p+sum of all correlations Variances and covariances : With sigma 2 i = variance of each item With sigma 2 T = variance of total scale (all p items) Variance of total scale With sigma ij = covariance between i and j

Q: What is an easy form of validation?

Content validity or face validity which means whether you can affirm that the items look like they capture the content you are hoping to measure .

4 important questions on Measuring customer satisfaction and factor analysis - Measurement theory

How can you cancel out the measurement errors in surveys?

By asking multiple questions on the same topic so that the scale reflects the basic nature of the concept.

What type of correlations do you expect when the survey is good?

High correlations between the grouped questions about one topic.

What index is used to capture reliability and how is it calculated?

Coefficient alpha. Ranges from 0-1 and best when above 0.7
Standardized:

With p = number of items in the scale (i.e. how many questions in the group)
With the denominator = p+sum of all correlations

Variances and covariances:

With sigma²_i = variance of each item
With sigma²_T = variance of total scale (all p items)

Variance of total scale

With sigma_ij = covariance between i and j

What is an easy form of validation?

Content validity or face validity which means whether you can affirm that the items look like they capture the content you are hoping to measure.