#1
Dear all,

I'm currently working on my thesis, which is an analysis of the effects of social cycling on cyclists' stress levels.
An experiment with a total of 21 participants was conducted. The Experiment was about a comparison between single cycling vs group cycling.
Single rides and group rides were performed while measuring the GSR (stress) data, in order to notice if there are any effects or changes on stress when riding in a group compared to riding alone.
Every participant has to participate in a single ride, followed then by group rides, changing the position each time ( riding in front, middle, in the back).

The plan was to use a simple t-test for analyzing (single vs group).
But we also noticed interesting effects between the stress values inside the groups depending on the position (front vs back, front vs middle..)

My supervisor meant that I should use analyze 2 factors using two-way ANOVA:
One factor would be cycling (single, group) and the other factor is position (front, middle, back) with stress values as the fixed factor.

I'm new to statistical analyses and currently using SPSS, and my question is: is the two-way ANOVA the right approach for my case? Because while adding the data i noticed 'single' will only have one position, while 'group' would have (front, mid, and back), meaning that group would be repeated many 3 times more than single. could that be a problem?

Apologies for the long question and if anything doesn't make sense.
 

fed2

Active Member
#2
i think ur right, you would have to run as a one-way layout, ie 4 treatments single and group (f,c,b). congratulations on outsmarting your supervisor. youll want to put this in terms of making it their idea though....

does their position change during the ride? (afraid to ask...)
 
#3
Hi Fed, thanks for your reply.

does their position change during the ride? (afraid to ask...)
The order doesn't change during the ride. A participant does a ride alone (single), then a ride with a group while riding in front, then another ride while cycling in the middle, and lastly another group ride while cycling in the back. So basically each participant participates in 4 rides.

i think ur right, you would have to run as a one-way layout, ie 4 treatments single and group (f,c,b).
I'll try the one-way layout, with single, front, middle, and back. did i get that right?

congratulations on outsmarting your supervisor. youll want to put this in terms of making it their idea though....
I didn't have the chance to ask or get more information from my supervisor, since he went on holiday. and I'm honestly running out of time and can't afford to wait for him to get back for more instructions..
 

fed2

Active Member
#4
Yes it is a one-way ANOVA as you describe with one-way layout, with single, front, middle, and back. But, very important, it is going to be a one way repeated measures design because of the same rider alone and in group.
The order doesn't change during the ride. A participant does a ride alone (single), then a ride with a group
This is extremely important to doing the analysis correctly. Sorry I did not notice before.

Also, hate to 'go there' given the time issues you have, but probably need to think about controlling for the 'ride id', ie the common effect of bein gon the same ride.

Well your experiment is sort of complicated really...
 
Last edited:

Karabiner

TS Contributor
#5
Tailoring tests to "interstingly-looking" patterns in the data, after eyballing data ,
produces dubious results. So what's all the discussion for? The p-values are distorted.

Moreover, the experimental design (each subject has exactely the same sequence)
confuses sequence effects with experimental condition. This harms the single/group
comparison and also the "position" comparisons.
 

fed2

Active Member
#6
Tailoring tests to "interstingly-looking" patterns in the data, after eyballing data ,
produces dubious results. So what's all the discussion for? The p-values are distort
this level of sanctimony regarding typeI errors probably isn't appropriate here. You were probably data-snooping as you wrote this...

Moreover, the experimental design (each subject has exactely the same sequence)
confuses sequence effects with experimental condition. This harms the single/group
comparison and also the "position" comparisons.
if they are all in same sequence, so be it. thats just a limitation of the research i guess. its not like you can go back and rerandomize.
 

Karabiner

TS Contributor
#7
I am afraid that I do not quite understand what you want to tell me.
To scan data and "test" some pattern found in the data is undistiguishable
from fraud. Regarding the second statement, I suppose we both agree
that the whole study is scientifically worthless, regarding sample size and
poor experimental design , but they have to pretend to do a study in order
to gain a bachelor degree or something like that. Nevertheless, it is
neccessary to point at the weaknesses, since maybe someone might
find this thread using a search engine.