How to calculate the overall mean of several Likert scale data?

#1
Hi, I have a set of questionnaire as follow:

Perceived Usefulness Rating(5-point Likert Scale)

PU01
PU02
PU03
...
I used SPSS to compute the mean of these data. The output as follow:

Descriptive Statistic N Minimum Maximum Mean Standard Deviation

PU01
PU02
PU03
Valid(listwise)

The result show that mean is compute for each PU01, PU02, PU03. This is not the result that I desired.

The question is how can I calculate the mean of the PU01, PU02, PU03 under a category which is Perceived Usefulness?

*English is not my native language.

Thank You.
 
P

parsec2011

Guest
#2
You need to give a value to your Likert-scale items. That is to say:

Strongly agree = 5
Agree = 4
Neither agree nor disagree =3
Disagree = 2
Strongly disagree = 1

Example:

Lets say respondent 1 answered disagree, respondent 2 answered strongly disagree, and respondent 3 answered neither agree nor disagree.

In this case, you have three "converted" values: 2, 1, 3. Its mean or average equals 2.

I believe it is more meaningful to use a different scale to 1,2,3,4,5 for your code. If you use multiples, 2.5, 5.0, 7.5, 10, you get somewhat of a more meaningful scale in my opinion. Although I am unsure whether there would be any serious statistical implications.

To learn more about this technique in SPSS, I recommend:

http://www.uni.edu/its/support/article/604

All the best,

Ramon
 
#4
If you're using a validated scale the original authors will have provided instructions on how it should be scored. If you're asking for advice on how to obtain a mean score across the items in SPSS then you can run the following syntax (adjusting for however many items in the scale):

Code:
compute PUmean = mean(PU01 to PU03).
exe.
 
Last edited:

SiBorg

New Member
#5
If you're using a Likert scale, consider using the median rather than the mean, as you're dealing with ordinal rather than continuous data.
 

bryangoodrich

Probably A Mammal
#6
I believe it is more meaningful to use a different scale to 1,2,3,4,5 for your code. If you use multiples, 2.5, 5.0, 7.5, 10, you get somewhat of a more meaningful scale in my opinion. Although I am unsure whether there would be any serious statistical implications.
Psychology isn't my area, but I'm curious to know what research there is to back that up or that discusses this issue? It's something I've wondered about. I know more ratings (say, 7-point scale) can give you a finer partition of the provided space (metric), so numerical values like means more closely match their real-number space counterparts, but I would like to find some articles that talk about using different numerical encodings and their benefits, such as your suggested multiples.
 

noetsi

Fortran must die
#7
I think in practice it would depend heavily on how people interpreted those codes. For an ordinal scaled variable I am not sure of the logic of assigning any particular scale more or less value. If you chose to interpret a likert scale as interval like (as is often done, but much debated) the mean of the later scale would be more confusing or less simple in any case.