# Interaction in time series analysis

I have three different physiological variables--heart rate, respiratory rate and blood oxygen saturation, each as a time series. I am trying to study the interaction between the variables as they progress towards a particular defined point in time. I tried the use of a Vector Autoregressive (VAR) model but unfortunately was not successful in trying to study the variables since I could never come up with a stable VAR model after many different iterations and crossing all the checks required for VAR modeling. I am wondering if I could get an idea or two on how I could study interactions between the variables in time as they progress to a defined point?

• The term "interaction" has a specific meaning in statistics, which does not seem related to what you are talking about. Can you clarify what you mean by the term & what you are after exactly? Jul 17, 2014 at 2:10
• True--point taken. Jul 17, 2014 at 2:20
• True--point taken.True--point taken. Apologies for the "not so technical" jargon. But the idea is to study how one variable and its subsequent values affect the other two variables and their values. For instance, I want to be able to show, for instance, that in the case of particular patient where all the 3 variables are collected, that decreasing heart rate also produces a decrease in the value of the blood oxygen saturation (or not)--as the case may be. More so, capturing the dynamics of the variables as they progress in time. Jul 17, 2014 at 2:27
• How many patients do you have these data for? How long are the time series? Are they the same for every patient? Are the measurements all at the same time points? Jul 17, 2014 at 2:32
• Yes the measurements are all at the same time points every 20 seconds for each of the variables--uniform time series. N = 200 Jul 17, 2014 at 2:56