Is it possible to use a statistical test on my population sample?

#1
Hello everyone,

I am currently doing a Research project and am unsure if its possible to assess significance. In my project, I have looked through the CT scans of 200 patients for certain findings (lets call them Feature A, B , C etc.). After totalling up the number of each type of finding, I wondered if its possible to assess whether, for example, finding Feature A in 45/200 patients is significant? Can I assess whether these findings are significant within my sample, or is it only possible by comparing against another sample population?

I would be grateful for any advice received! Thanks for your time.
 

Karabiner

TS Contributor
#2
You can define a 22.5% as "significant" (important, relevant, stiking...) within your area of study, if you have reason to do so.

But if you have "statistically significant" in mind, then you need to specify which comparison(s) you want to make. This can
for example be a comparison between groups, or between features A and B and C, or whether A is significantly higher (or
lower) than a theoretically meaningful rate.

With kind regards

Karabiner

BTW, "sample population" is not very useful a term, either it is a population, or it is a sample from that population.
Not both at the same time.
 
#3
You can define a 22.5% as "significant" (important, relevant, stiking...) within your area of study, if you have reason to do so.

But if you have "statistically significant" in mind, then you need to specify which comparison(s) you want to make. This can
for example be a comparison between groups, or between features A and B and C, or whether A is significantly higher (or
lower) than a theoretically meaningful rate.

With kind regards

Karabiner

BTW, "sample population" is not very useful a term, either it is a population, or it is a sample from that population.
Not both at the same time.
Appreciate the assistance! So yes unfortunately I cannot compare within the group. The only comparisons that can be made is between features A, B, C etc.

Thanks for clarifying
 

hlsmith

Less is more. Stay pure. Stay poor.
#4
What do you hope to do with the results? Also, if you did suspect the association before running comparisons, you risk for false discovery will be high if not addressed!
 

Karabiner

TS Contributor
#6
If you mean inferential statistics (tests of significance), then you do not need any.
These tests are used for making conclusions about population parameters (e.g.
whether a difference between means might be zero in the ppopulation, or
whether a correlation coefficient might be zero in a population, or whether
a difference between proportions might be zero in the populuation), based on
observations which were sampled from that population. If you have the whole
population already at hand, then there is no need for inference (from samples
to populations) and therefore, no need for a test.

With kind regards

Karabiner