# Setting PyMC model with two different time series data

I've been working with PyMC for a bit, and am stuck on this one. I see example on fitting time series, in the tutorial and others like:

http://lighthouseinthesky.blogspot.com/2011/10/curve-fitting-part-5-pymc.html

and

http://healthyalgorithms.com/2010/10/19/mcmc-in-python-how-to-stick-a-statistical-model-on-a-system-dynamics-model-in-pymc/

However, in each case, we either have a single variable, or data for two or more variables for the same time points (so the vectors are all of the same length). What happens if my data is like:

t=[1,2,3]
x=[.5,.8,.2]


and

t=[1.5,2.5]
y=[.1,.5]


with some simple model, like

x=a*t+b
y=c*t+d


the model isn't important. the point is that the "observed" data have different time points. I haven't been able to find any example like this, without perfectly aligned data. Are there any? Is there a simple example that someone could point out?

thanks!

 Brian Blais

• are x and y related somehow? Otherwise this problem is only two simpler problems. – Cam.Davidson.Pilon Apr 3 '13 at 14:20

I may be misunderstanding what you are after, but if you are modeling time t explicitly, it should not matter that they are not aligned, does it? You might have a peek in Chapter 11 of Gelman and Hill for examples of Bayesian models for longitudinal data.