Hypothesis test on ordinal data

[ultimatejester]

Can i use the same the techniques used for nominal data on the ordinal data. I can't seem to find help on hypothesis test for ordinal data. This question contains categories (customer satisfaction) and they are in order so the data must be ordinal but they haven't taught us how to conduct a test on the ordinal data.


btw why isn't there any info on conducting a test on ordinal data anywhere. Not even google can find anything. Is it even possible to conduct an inference on nominal data.

Thank you.

 

[EnumaElish]

You can; with caveats. Try googling chi-square test (http://en.wikipedia.org/wiki/Chi-square_test)

 

[ultimatejester]

Thank you for replying(much sooner than i expected :)

I am familiar with chi distribution but i don't think it would work here since the population std. deviation and other attributes are unkown and because the data is ordinal i can't use basic techniques to figure them out either. All i have is the sample size which is 700 customers and their satisfaction rating with the company

1 = highly satisfied
2 = satisfied
3 = neutral
4 = dissatisfied
5 = highly dissatisfied

I will make another post right after this to elaborate further.

 

[ultimatejester]

I could do proportions and use test statistic for proportions but the neutral category causes problem because that splits the customer base into categories.

-Satisfied
-Neutral
-Dissatisfied.

 

[CRGreathouse]

You could use binary variables:

Is the customer highly satisfied?
Is the customer at least satisfied?
Is the customer at least neutral?
Is the customer at least dissatisfied?

 

[EnumaElish]

Thank you for replying(much sooner than i expected :)

I am familiar with chi distribution but i don't think it would work here

Yes it would; see http://en.wikipedia.org/wiki/Pearson's_chi-square_test

You can also define binary "dummy" variables as CRGreathouse has suggested and test them individually or jointly.

 

[ultimatejester]

i got it figured out. Thanks for the help. It is greatly appreciated.

 

[msdinh]

hey

this may sound weird but i have the same question well it seems the same.. and I am stuck I am just wondering how you did it? did u use the population proportion?

 

[ultimatejester]

Yes, i used proportions. You can change the code for success from 1,2,3,4,5 but the one you are really interested in is the "highly satisfied" or 1. I did a one tail test since the company wants to know where its is meeting the required 95% level. To get you started

H0: u=95%
H1: U<95%

Good Luck

 

[bigdaddyindia]

satisfaction question

so would you recommend using the t-test, since the std. dev. is unknown? also, wat would the null and alternative hypotheses be? thanks a lot!

 

[ultimatejester]

The null and alternative hypothesis are mentioned in my above post. You can't use the t-test. All you are allowed to do on data in count the occurences. Use proportions. Post here if you need more help.

Thanks.

Adil.A

 

[LaurenA915]

Hi,

I was wondering if u could post how u did that question.

I'm having the same problem and i still can't figure it out :(

Any help would be really appreciated! Thanks!

Dona.