simple count - type of data

#1
Hi everyone
If I have data which is in the form of a count - in this case number of words recalled under two different conditions - is this nominal, ordinal or interval data?

There is a total of 50 words which could be recalled so the data range is 0-50. I thought this was ratio data but my tutor says as it is a simple count it is nominal and I have to convert it to a percentage before running parametric tests. I don't understand why this is the case.

Can anyone help?
 

gianmarco

TS Contributor
#2
In my opinion, you have two categorical variables: words recalled (with 50 levels, 1 for each word) and condition type (with 2 levels).
Besides understanding what type of data you are dealing with, it would help knowing your research question or analytical goal.

Best
Gm
 

AngleWyrm

Active Member
#3
Let's say I'm a test subject, and I recall 15 words. That information alone isn't very informative until we also include the fact there are 50 words. Then we have a proportion: 15 out of 50 words = 15/50 = 3/10 = 30% of the total words.
 
#4
Just a simple t-test between the two groups to see the difference in recall between them. But I am told the data as it stands (number of words recalled out of 50) is "nominal level" and therefore does not meet parametric assumptions so it must be converted into a percentage before running the t-test. This did not fit with my prior understanding of nominal, ordinal, interval and ratio data so I am trying to understand!

In my opinion, you have two categorical variables: words recalled (with 50 levels, 1 for each word) and condition type (with 2 levels).
Besides understanding what type of data you are dealing with, it would help knowing your research question or analytical goal.

Best
Gm
 
#5
Let's say I'm a test subject, and I recall 15 words. That information alone isn't very informative until we also include the fact there are 50 words. Then we have a proportion: 15 out of 50 words = 15/50 = 3/10 = 30% of the total words.
Thanks for your reply - I understand what you are saying but I am not entirely sure how it relates to parametric testing. I have been told to convert the "nominal level data" (number of words recalled out of 50) into parametric data in order to run a t-test. Do you know why I can't just run a t-test on the data as it is? And why does making it a % make it parametric? In effect, the figures are just doubled in this case e.g. 25 words recalled out of 50 is 50%. I appreciate any explanation as I am struggling to get this.
 

hlsmith

Less is more. Stay pure. Stay poor.
#6
What is the study purpose? Tell us the goal of the project - so we aren't the blind leading the blind.

IMHO - this actually seems like count data that should be examined using Poisson regression.
 

Miner

TS Contributor
#7
Counts are counts, not nominal. You can have binary, nominal or ordinal 'categories' with counts for each category. Counts would normally follow a Poisson distribution, but with 50 possible levels, it becomes pseudo-continuous data and you should be able to use a t-test. You could also convert it to a percentage and use a test for 2 proportions or an Analysis of Means (ANOM) for binomial data, or leave it as count data and perform an ANOM for Poisson data.