One Sample T-Test with Likert Scale Questionnaire

#1
Hi all,

I am doing a research and will study the how user think
about shopping online.

In my questionnaire, I have 3 questions - using Likert Scale
1 - Strongly Agree, 2 Agree, 3 Nuetral, 4 - Disagree, 5 - Strongly Disagree.

My questionnaire is as follow.


1. I find that Internet is secured for shopping
2. My general intention to shop online is high.
3. I find that shopping online is convenient.


The Hyphothesis that I want to test is

H1 - Users has confidence to shop online.

-------------------------------------------------

Please advise the basic procedure and step to test my H1
using, say, for example, one-sample t-Test (which
is the easiest, I think)

Many thanks.


Regards,
Very Worried
 

TheEcologist

Global Moderator
#2
Hi all,

I am doing a research and will study the how user think
about shopping online.


Regards,
Very Worried
I don't think you have thought your design through all the way. Using statistics we mostly 'test' for difference between groups (as with a T-test, Anova ect; Chinese men are taller than Korean men ect) or you try to relate one variable to another (temperature and local ice-cream sales).

What are you testing your hypothesis against?

I'm missing this:

Users have confidence to shop on line compared to what?

or

User have confidence to shop on line because there mean score is higher than 3.5 on a Likert scale. > 3.5 is considered confident because.....

Once you have defined the above well then you have a working hypothesis and this can be tested.
 
#3
Hi,

Ok. Let me rephrase my question. I wasn't specific in my previous
question.

My study is on Consumer's Attitute Towards Online Shopping.

One of the factors, based on my study (literature review) is "Incentive".

And now, in my research methodology, one of the Hyphothesis I will
test is

H1 - Incentive will encourage users to shop online.

(note: Incentive may be online coupon, discount etc).

So, in my questionnaire design, I need to check if this H1 is true.

I have ONE question in my questionnairre that relate to "Incentive".
(using Likert scale)
1- Strongly agree, 2,3,4 ...... 5 Strongly disagree.

e.g.

Q23 I find that reward will encourage me to shop online. (1,2,3,4,5)

Note: Here, I only have "one" question that "relate" to "incentive" and
online shopping.

So, my question, is ...

a. Can I use one sample t-test to test this ?

Correct me if I am wrong. I believe, from the test, I will have
something like, eventually, p < 0.05 or p > 0.05 and there,
I can either reject or accept H1.

Other related questions maybe be

i. For Likert scale, we have 1 - Strongly Agree, 2 Agree etc.

Do you normally, assign a number to the scale so that,
we can do the calculation of Mean?

Say

1 (Stronly agree) - 4
2 (Agree) - 3
3 (Neutral) - 2
4 (Disagree) - 1
5 (Strongly Disagree) - 0

And to get the mean of the sample, since
"Strongly Agree" and "Agree" will be accepted and considered
positive (for accepting online shopping when reward is given),
do I have to calculate the Mean for both (1 strongly agree) and
(2 Agree) and then average the Mean and divide by 2 ?
( may be a silly question haha)


ii. Once I have the Mean, next, I will need to define the
Hypo Test Value.

Now, normally, how do define the Hypo Test value?
Are there any rules? Normally, do we get it from the
Mean of people who says "Neutral" ?

iii. And, I have more than 1 Questions, e.g. add one more question
and users about Incentive and Online shopping, say now, I have


Q23 I find that reward will encourage me to shop online. (1,2,3,4,5)
Q24 ************************************************x (1,2,3,4,5)

now, since these 2 questions relate to "Incentive" factor,

for the mean calculation, I have 2 "Means". Should I, again,
average them up in order to get an average Mean ?

(maybe another silly question. ha)

Or for 2 or more question, is one-sample t-test is not applicable?



Many thanks for your advice.


Regards,
Still Very Worried
 

CowboyBear

Super Moderator
#4
If you want to see if there's an association between incentive and online shopping, you're looking for a correlation, not a one-sample t-test.

It sounds like you really need to hit your textbooks and get a basic idea of what test is appropriate and how it's used - it isn't really fair to throw up every question you think of, and expect others to answer them for you.

If you really can't find the answer by doing some googling and reading, put single specific questions on the boards that show you've made some attempt to work the answer out for yourself.

By the way, you might need to get your questionnaire proofread - items like "I find that Internet is secured for shopping" "I find that reward will encourage me to shop online" are nowhere near grammatically correct (apologies if you're translating from a questionnaire in another language).
 
#5
Don't answer then. I don't need any lecture from you. If you don't think
of helping, just keep quiet and get lost.

I am asking courteously for advice and NOT answer.
 
Last edited:
#6
The previous message was meant to ... CoyboyBear.

By the way, one-sample t-test was used by many researchers (they published
some papers in fact) and it was used by them to test similiar cases.

TheEcologist's advice seemed to make more sense. I got some idea from
your advice and bought SPSS software to try out the one sample t-test.

I basically, use the "Compute Variable" to resolve the "multiple questions"
to one variable issue.

Thanks.
 

TheEcologist

Global Moderator
#7
If you want to see if there's an association between incentive and online shopping, you're looking for a correlation, not a one-sample t-test.

It sounds like you really need to hit your textbooks and get a basic idea of what test is appropriate and how it's used - it isn't really fair to throw up every question you think of, and expect others to answer them for you.
I tend to agree with CowboyBear; I dont see why you would want to use a t-test (please read:http://www.socialresearchmethods.net/kb/stat_t.php).

Also check this useful site, look at the tests and why they are used and assess critically where your problem fits in:

http://udel.edu/~mcdonald/statbigchart.html
 
#8
Hi all,

I am doing a research and will study the how user think
about shopping online.

In my questionnaire, I have 3 questions - using Likert Scale
1 - Strongly Agree, 2 Agree, 3 Nuetral, 4 - Disagree, 5 - Strongly Disagree.

My questionnaire is as follow.


1. I find that Internet is secured for shopping
2. My general intention to shop online is high.
3. I find that shopping online is convenient.


The Hyphothesis that I want to test is

H1 - Users has confidence to shop online.

-------------------------------------------------

Please advise the basic procedure and step to test my H1
using, say, for example, one-sample t-Test (which
is the easiest, I think)

Many thanks.


Regards,
Very Worried
I have done a research and have used Likert type scale. What type of hypothetical tests am i going use?
 
#10
Hi Dalhatu, why did you go for the Likert scale?
I don't think you have thought your design through all the way. Using statistics we mostly 'test' for difference between groups
(as with a T-test, Anova ect; Chinese men are taller than Korean men ect) or you try to relate one variable to another (temperature and local ice-cream sales).

What are you testing your hypothesis against?

I'm missing this:

Users have confidence to shop on line compared to what?

or

User have confidence to shop on line because there mean score is higher than also get phenq deals 3.5 on a Likert scale. > 3.5 is considered confident because.....

Once you have defined the above well then you have a working hypothesis and this can be tested.
Because Likert scale is the best scale to gauge user thoughts, in my humble opinion. What other scale do you suggest?
 
Last edited: