1
$\begingroup$

I read through some literature online and did some investigating on my data and the type of scales my data represents. In my data set, I have my independent variable called ASA status (this is a categorical variable, ordinaly ranked, with 2+ groups or levels), and my dependent variable which is the number of events, or count, based on a patient's ASA status. What I would like to do is to use a statistical test to determine whether an increase in the patient's ASA status and the number of events that patient has during their procedure are statistically significant. So basically I want to prove that a higher ASA status would result in a higher count of events for each patient. My data looks something like this (pairing each patient's ASA Status with the # of events they had per case) with over 100 samples:

ASA Status & Number of Events: (2, 1), (3e, 2), (4, 2), (1e, 1), etc...

(just an FYI the 'e' is not a typo next to the numbers; those denote a specific ASA Status, which is why I said my IV are categorical).

What I noticed for the one-way ANOVA and the Kruskal Wallis tests is that both tests allow for interval-type dependent variables. I imagine that this has to do with parametric versus non-parametric assumptions, correct? The problem I'm having is determining whether they fit one or the other, so here is the structure of my data in a nutshell:

  • The data was selected with the intent of only using cases where at least 1 event was present in a case.
  • There are a total of 12 levels for the IV for this study (some of which have 0 data points). The ordinal ranking of this IV is as follows: [most severe] 6e > 6 > 5e > 5 > 4e > 4 > 3e> 3 > 2e > 2 > 1e > 1 [least severe]
  • Number of events is strictly a whole number (cannot have fractions of an event)
  • Some of the ASA Status levels have very few cases (one of them has only 1 case, and some of them have 0 cases).

So I'm pretty lost at this point in determining whether I should use one-way ANOVA or Kruskal Wallis. Or is there another test that I did not consider that I should use?

Any assistance with my question would be a great help.

$\endgroup$
  • $\begingroup$ ANOVA can be used if your dependent variable follows a normal or near normal distribution, and Kruskal-Wallis test can be used otherwise. $\endgroup$ – prashanth Aug 22 '18 at 11:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.