Comparing "groups" - which test should I use

#1
Hello everyone, thanks for being on my topic.

I am doing a research on parkinson's disease. I have several things to do, but now, i'm stuck with statistics.

This research use cinematic parameters of hands movements to create new clues for doctors to get a more precocious "diagnosis" (we use ROC curve).

So, i have X patients with Parkinsons disease, i see them twice, and i have Y healthy subjets, seen twice also.

Now I want to compare the peak of acceleration of my patients those way:

The dominant hand to the more affected hand

The dominant hand to the less affected hand

The non dominant hand to the more affected hand

The non dominant hand to the less affected hand

But i don't know which test to use. I checked http://abacus.bates.edu/~ganderso/biology/resources/stats_flow_chart_v2014.pdf and https://www.statsflowchart.co.uk/ but it's still a wall to me.

I use Excel or JAMOVI. But if you think i shoud use something else. Do not hesitate.

I don't know if i am crystal clear, if you have questions in order to help me xD Feel free to ask of course !

Thanks everyone.
 

Karabiner

TS Contributor
#2
So, i have X patients with Parkinsons disease, i see them twice, and i have Y healthy subjets, seen twice also.
Now I want to compare the peak of acceleration of my patients those way:
That is, you only want to analyse your patients, not the healthy subjects?
The dominant hand to the more affected hand
The dominant hand to the less affected hand
The non dominant hand to the more affected hand
The non dominant hand to the less affected hand
Admittedly, I do not know what you mean.
But i don't know which test to use.
Mind that statistical tests of significance can only state such things as "based on the results,
we reject the null hypothesis that the mean difference is exactely 0.00000... in the population
from which the data were sampled; instead, we assume that the mean difference is larger
than 0.000000..." For diagnostic purposes, such conclusions might not be extremely useful.

How large is your sample size, by the way?

With kind regards

Karabiner
 
#3
Hello Karabiner, thanks for your reply.

Well, yes, with this question i only want to analyse the affected patients. That's right.
I use the healthy subjet to create a ROC curve.

Admittedly, I do not know what you mean.
With parkinsons disease, your dominant hand is more "parkinsonian" compare to your not dominant hand. So there is a more and a less affected hand.
My thesis director asked me to compare this 4 ways, with cross comparison.
I do understand it's kind of odd in some cases, i told him but he seems to be sure of him so... i ask because maybe i'm dumb and misunderstand something here.

Mind that statistical tests of significance can only state such things as "based on the results,
we reject the null hypothesis that the mean difference is exactely 0.00000... in the population
from which the data were sampled; instead, we assume that the mean difference is larger
than 0.000000..." For diagnostic purposes, such conclusions might not be extremely useful.
Yes that's true.

How large is your sample size, by the way?
I have around 30 subjets.

Kind regards,
John
 

hlsmith

Less is more. Stay pure. Stay poor.
#4
The four comparisons don't make sense to me either and I get the general premise. You have 20 Parkinson patients and have two measurements on their hands. You want to see if the decrease in functionality of my dominant left hand versus my dominant left hand post. and all iterations. Why distinguish between left and right hands, if that is what you are doing? Also, is the time between the two measurement the exact same between all patients or is there variability?
 

Karabiner

TS Contributor
#5
If you want to compare some interval scaled variable, measured at twice at the same subjects,
you can use dependent samples t-test or Wilcoxon signed rank test.

WWith kind regards

Karabiner
 
#6
The four comparisons don't make sense to me either and I get the general premise. You have 20 Parkinson patients and have two measurements on their hands. You want to see if the decrease in functionality of my dominant left hand versus my dominant left hand post. and all iterations. Why distinguish between left and right hands, if that is what you are doing? Also, is the time between the two measurement the exact same between all patients or is there variability?
The time between two measurement is minimum 7 days for all participants. The appointments are schedule according their possibility for the study, so there is variability, with some i have exactly 7 days, with others i have 3 weeks between.
I want to compare the decrease or increase of the peak of accelerations the IMU recorded.
I did not understood very well what you mean by "my dominant left hand versus my dominant left hand post"
 
Last edited:
#7
If you want to compare some interval scaled variable, measured at twice at the same subjects,
you can use dependent samples t-test or Wilcoxon signed rank test.

WWith kind regards

Karabiner
Can i use those tests with : measure type "continuous" and data type "decimal". That's what JAMOVI elected. I have that kind of results for exemple for my peak acceleration of movement:
"26646.399
2125863.905
0242489.123
4231992.972
0626847.679
2829318.834
7944060.454
..."

Thanks very much to you both :)