Choosing a statistical test.

AMay

New Member
#1
Hi,
I am not very confident with stats but am needing help with analysing my data.
I have observational data gained from observing marmosets over a period of five weeks. I am aiming to examine if there is any significant increase in time spent on certain behaviours over the course of these 5 weeks. I have two marmoset subjects. For each I have the total percentage of time spent on 8 different behaviours (stationary, grooming others, grooming self, playing, resting, locomotion, foraging natural and feeding on provisions) for each week (week 1,2,3,4,5). Am I correct in thinking that my independent variable is ‘percentage of time spent’ and my dependent variables is ‘week number’? Also Which test would be correct to use for this data?.
Any help would be greatly appreciated.
Thankyou.
 
#2
I am aiming to examine if there is any significant increase in time spent on certain behaviours
Then the dependent variable is "the total percentage of time spent" on different activities. So you want to know if there is an increase in the percentage of time spent on e.g. playing as weeks go by.

It is the percentages that you want to explain. And what is going to explain that is the independent variable = the explanatory variable or the x-variable.

Maybe the percentages can be modelled as Dirichlet distribution, (with 8 different activiies, 8 different percentages) where all percentages sum to 1 (i.e. 100%)

(Search for the Dirichlet distribution and software for it.)
 

AMay

New Member
#3
Many thanks for your response. I was hoping to use R studio to analyse this data and want to make it as simple as possible as I have limited ability. Would a one-way within anova/repeated measures test work with this data?
Thanks.
 
#4
Would a one-way within anova/repeated measures test work with this data?
That sound kind of crude. If the proportion is larger than 0.2 and less than 0.8 it might be OK. But since you you have 8 classes so some percentages might be small.

The beta distribution is a special case for the Dirichlet distribution. (So take one proportion at a time.) Maybe this or this can help.