Question about data scale

SiaDX

New Member
#1
Hello,
I have a survey with the following question:
"Which price do you think is fair for Product A?"
  1. Less than 14.95: (Enter a value here)
  2. 14.95
  3. 19.95
  4. 24.95
  5. 29.95
  6. 34.95
  7. More than 34.95: (Enter a value here)
What type of data do i get from this and which analysis can I conduct? I am assuming it is on a ratio scale which means I can do a t-test after however the categories are not equally spaced as people can enter their own preferred prices at the first and last category. Someone told me this is an ordinal scale which I dont understand as I have discrete numbers here and not categories. I am really confused and I really need to know what type of data this is in order to proceed with the statistical analysis of the results. Thanks in advance!
 

katxt

Well-Known Member
#4
But probably that's not the real question. There is a mention of a t test. So are the subjects categorized into different groups and we want to know if the means are different? Or is there a true known value and SiaDx wants to know if the subjects are different from this? Or perhaps compare one product with another? Or what?
 
Last edited:

SiaDX

New Member
#5
But probably that's not the real question. There is a mention of a t test. So are the subjects categorized into different groups and we want to know if the means are different? Or is there a true known value and SiaDx wants to know if the subjects are different from this? Or perhaps compare one product with another? Or what?
The question above is posted to german and swiss customers and I want to investigate whether there is a difference in what these two groups consider to be a "fair" price for a certain product.
 

katxt

Well-Known Member
#6
OK. So it sounds like a two sample test situation. Which actual test to use depends on how many respondents you have. If you have a lot of data then a t test using the actual data will give reasonable results even if the data isn't really normal. Perhaps a more informative approach would be to find a confidence interval for the difference.
A less common method which gets round the problem of the ends and non normality would be a resampling or a permutation test.
Multiple p values may be a further complication if you are testing multiple products.
As you have found, there are subtle difficulties in your current plan which could perhaps been avoided with a different survey design.
Who is the report going to, and what sort of thing are they hoping for?
 

Karabiner

TS Contributor
#7
The question above is posted to german and swiss customers and I want to investigate whether there is a difference in what these two groups consider to be a "fair" price for a certain product.
A comparison between means using independent sample t-test might
give reasonable results here, if sample size is not very small (at least n > 30).
In addition, you could compare the variances betweens groups, i.e. you can
analyse whether Swiss and German customers differ with regard to
degree of consensus within their respective group.

If you want to account for the fact that the data might be considered ordinal,
you can use median test (for the comparison of median values between groups),
or Wilcoxon rank sum test / Mann Whitney U-test (analysing whether the
values in one group tend to be higher than in the other group).

With kind regards

Karabiner