Independent samples T Test or Chi-Square

ynos

New Member
#1
Hello!



I'm writing masters degree and cannot decide which method to use. :confused:




Hypothesis 1:


The image of Englans as a tourist destination is different with respondents who have already visited this country as opposed to respondents who have never been to England.




Have you ever been to England?

-yes (1)
-no (2)



Please evaluate your overall image of England as a tourist destination:

1 2 3 4 5 6 7


1 - very bad

7 - very good



If I decide to use "Independent Samples T Test", overal image should be normaly distributed, as I remember. I would test mean defference...


If I decide to use "Chi-Square test", there are more than 20% expected cells have count less than 5.



Here is actual data in SPSS:

SPSS data sheet


I noticed that standard p-level reported with a t-test are smaller than p-levels when I use Chi-Square. So I can get to different conclusion (rejecting/accepting Hyphotesis) with different tests.



I'm really confused right now. Thank you for all your help.
 
#2
Hi,

The scales which you have used will be treated as ordinal scales.For ordinal scales measurement you can use Mann-Whitney Test (t-test can be used if you treat your scales as interval scale).chisquare test can be used when you measure it using a binary variable

Regards
Re.Vinaitheerthan
 

ynos

New Member
#3
Isn't it INTERVAL scale?


You can express difference --> Y (a) - Y (b) = d

And you are able to calculate Mean and Standard Deviation from it.



Thank you


n = 100



Although Likert scales are really ordinal scales (definition of ordinal scale), they are often treated as interval scales. By treating this type of agreement scale or attitudinal measurement as interval, researchers can calculate mean scores which can then be compared. For instance, the level of agreement for men was 3.5 compared to 4.1 for women, or it was 3.3 for first time visitors compared to 2.8 for repeat visitors.
 
Last edited:
#4
You are using categories for people to answer your questions (yes/no and very bad to very good) and not likert scaling to several questions to create a composite total.

rs