An Analyst is working on developing an appropriate scale measure of the Loyalty of a company's customers.

After some initial work the SPSS Cronbach Alpha output is as follows (please see attached). Comment on the analyst's efforts and advise what changes you would make to improve the scale if necessary.

My Answer:

The analyst should consider checking the missing data to determine whether respondents weren't asked certain questions (there were errors with the questionnaire routing), did respondents simply not answer and why they didn't (question could be irrelevant and there is incorrect questionnaire routing ) or data was not recorded due to error while recording or editing data. Once the error is determined then the analyst should remedy the situation.

The negative Cronbach's alpha is indicative of violating reliability model assumptions

(i) Coding errors- the analyst should check whether they forgot to recode reverse scored items

(ii) Sampling error- the analyst should check small sample size umber of items can sometimes lead to a negative Cronbach's alpha (due to negative average covariance)

(iii) Incorrect set of items selected to measure loyalty- the analyst should run a factor analysis to ensure that the items are correct

(iv) Incorrect type of questions- the analyst should reduce the number of open ended questions or reconstruct them to a Likert scale