Total score issue

[E. Bernardo]

Dear Everyone,

 

We are developing a survey instrument that will measure the consumer perception of brand.  The instrument has 6 constructs and each contruct contains 8 items which use 10-point likert scale response format.  We want to have score for each construct then the global score..  We would be using the total score; thus, the score for each construct ranges from 8 to 64 points while the global score ranges from 48 to 384.  We want your thoughs about the following:

 

1.  What is the weakness of a 10-point Likert scale response format over the other point system?  It is a good practice to use 10-point system?

2. What is the scoring practices? What can you suggest, do we use total score or mean score?  

3.  For total score, can we use Cronbach alpha for the reliability test?  Can you please suggest an alternative for Alpha, if any?

 

Thank you in advance.

 

Eins

 

---

Yahoo! Mail Now Faster and Cleaner. Experience it today!

[Dan Riner]

1.       What is you reason for using a 10-point scale?  This is often done to try and get some additional variance out of a measure with traditionally low variability.  I did some work on this with a job performance measure a few years ago and found only very slight increases in variance when going from a 5-point to a 10-point scale.  If I recall, there are a few published studies that concur with my experience.  On the downside, you can get odd response patterns with larger scales such as respondents only using 1, 5, and 10 or only using the top or bottom of the scale.  I’d say stick with a 5 or 7-point scale unless you absolutely must squeeze out every last drop of variance you can get.

 

2.       For most purposes total score and mean score are going to give you the same information.  I’d be careful about using a total score across all your constructs unless they are quite strongly related empirically and theoretically.  Do you plan on running a CFA with consumer perception as a higher order factor of your 6 constructs?

 

3.       Your global internal consistency is going to be affected by how highly correlated your constructs are.  Did you or will you be running an EFA to help you select items for your scales?  If you use an orthogonal rotation your global internal consistency may be a little disappointing.  If you have a good theoretical reason for your constructs to be correlated you’ll probably want to use an oblique rotation.

 

 

Dan

 

Dan Riner

Statistician

Office of Institutional Research

US Naval Academy

[hidden email]

 

FOR OFFICIAL USE ONLY - PRIVACY SENSITIVE

Any misuse of unauthorized disclosure can result in both civil and criminal penalties.

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Eins Bernardo
Sent: Tuesday, January 19, 2010 8:45 PM
To: [hidden email]
Subject: Total score issue

 

Dear Everyone,

 

We are developing a survey instrument that will measure the consumer perception of brand.  The instrument has 6 constructs and each contruct contains 8 items which use 10-point likert scale response format.  We want to have score for each construct then the global score..  We would be using the total score; thus, the score for each construct ranges from 8 to 64 points while the global score ranges from 48 to 384.  We want your thoughs about the following:

 

1.  What is the weakness of a 10-point Likert scale response format over the other point system?  It is a good practice to use 10-point system?

2. What is the scoring practices? What can you suggest, do we use total score or mean score?  

3.  For total score, can we use Cronbach alpha for the reliability test?  Can you please suggest an alternative for Alpha, if any?

 

Thank you in advance.

 

Eins

 

 

---

Yahoo! Mail Now Faster and Cleaner. Experience it today!