Multiple measurement time series

Tass

New Member
#1
Hi dear helper,

I have a collection of monitoring data which I want to analyze as to significant differences over time.
I have monitored 2 groups of people over 16 months as to how often they feed their livestock per day. Both groups are of different size and each individual has a different number of livestock, though all animals of one owner are generally treated the same way. Not all people have livestock throughout the whole monitoring period though.
So I might have a total N of 50 people and 600 animals but in the first months I have 28 and 22 people, in the second I have 27 and 20 etc, with the animal numbers also changing as animals enter or exit.
However, on average I get differences in the feeding times each month and over all. I want to know now if these differences are significant.

What is the most appropriate way of doing this?

Mind you, in this case the data is obviously ratio data, but I also have other variables (like feed quantity) that are interval data or even nominal data when it comes to management and it would be perfect if the statistical approach could be applied to all of them. If not possible, I am already very grateful for a hint how to solve the above specified question.

I posted this earlier at another forum but got no reply, so hope to be more successful here. :rolleyes:

Best regards,
Tass.
 

noetsi

Fortran must die
#2
I think repeated measure ANOVA is going to be the simplest way. I think, although I am not sure, that you will have to address individuals leaving the system (which is what happens when they have no livestock given your analysis). I don't think you can simply replace some people with others or at least that is not the way I have seen this analysis done. That is for a controlled study. In practice I assume both ANOVA and Regression accept using different units over time in generating simple correlation (which is what most regression is in practice - not sure about ANOVA given how it is commonly utilized).

This assumes your dependent variable is interval (it is not clear to me what your dependent variable is - your predictors don't matter in this regard).
 

Tass

New Member
#3
Hey, thanks a lot, that is a good start for me. As to the dependent variable I am not sure either. I ant to compare 2 data sets after all. :confused:

Any hint is appreciated but I will think about it a bit more.

Thanks a lot.
T.