The Student Room Group

What stats test can I do for this? (Medicine related)

I am doing this dead module and I got a mock for it tomorrow. What stats test can I use to analyse data where:
Patients with a certain condition are given 2 sessions of treatment, after each session they are given a form to fill out which gives them a score. However pre-treatment scores are also recorded so there are 3 sets of data for each patient overall. The score should go down after each session. I am only using 10 patients (approved by my examiners).
I think the data is nonparametric and was thinking of doing a Kruskal-Wallis test but the examples online don't match up to what I'm doing it for.
Please help ASAP.

My data is discrete, dependent, 1 independent variable, hard to say whether it is parametric or not as of yet but I'm assuming not yet. I am thinking of doing a one-way anova, if data turns out to be parametric, or Krusalis-Willis Test.
(edited 2 years ago)
Reply 1
bump
Maths forum would be better for this.

So you've got continuous (?), dependent, ?normally distributed (you seem to be assuming not) data. Unfortunately that's as far as my stats knowledge goes :tongue:
Hi, I've moved your thread to the maths study help forum; please add as much information as you can about the question and also include what work you have done on it so far, and what your understanding of it is (as well as what level this is for) so the study helpers can provide advice :smile:
Original post by humayun.h
I am doing this dead module and I got a mock for it tomorrow. What stats test can I use to analyse data where:
Patients with a certain condition are given 2 sessions of treatment, after each session they are given a form to fill out which gives them a score. However pre-treatment scores are also recorded so there are 3 sets of data for each patient overall. The score should go down after each session. I am only using 10 patients (approved by my examiners).
I think the data is nonparametric and was thinking of doing a Kruskal-Wallis test but the examples online don't match up to what I'm doing it for.
Please help ASAP.

My data is discrete, dependent, 1 independent variable, hard to say whether it is parametric or not as of yet but I'm assuming not yet. I am thinking of doing a one-way anova, if data turns out to be parametric, or Krusalis-Willis Test.

Repeated measures ANOVA - and if your data is horribly distributed, either transform it until it is nice, or use something like the Friedman test (not something I have ever used!)
Reply 5
Original post by Gregorius
Repeated measures ANOVA - and if your data is horribly distributed, either transform it until it is nice, or use something like the Friedman test (not something I have ever used!)

Is the friedman test a test for 2 independent variables, whereas I only have one?
Original post by humayun.h
Is the friedman test a test for 2 independent variables, whereas I only have one?

But you're wanting to compare the score at the end to the score at the beginning aren't you? Otherwise what are you trying to do with this stats test?

(although surely the variable is not independent)
(edited 2 years ago)
What do you think the independent variable is? Is there a control group?

This looks like a one-way repeated measures ANOVA to me. Remember that ANOVA does not assume a normal distribution of the data, but a normal distribution of the residuals. It is also robust to violations of that assumption.

The Friedman test is a non-parametric alternative, but it has its own assumptions that mean it is questionable for time course data, apparently*. Full understanding is over my pay grade.

*Noguchi, K., Gel, Y., Brunner, E. , Konietschke, F. (2012). nparLD: An R Software Package for the Nonparametric Analysis of Longitudinal Data in Factorial Experiments. Journal of Statistical Software 50, Iss. 12.

Quick Reply

Latest