AN ANALYSIS OF A BIVARIATE TIME SERIES IN WHICH THE COMPONENTS ARE SAMPLED AT

It is desired to express the relationship between the components of a bivariate time series. What is unusual is that the components are observed at different times and that the observation times are irregularly distributed. The problem of different sampling times is dealt with by interpolating the values of the dependent series to the times of the independent. This is followed by nonparametric regression to estimate the relationship. The research is motivated by data collected at a station along the Solim?es River in central Brazil and also at a second station along a branch of the Solim?es. Of interest to geographers is the possible change in the proportion of the Solim?es waters entering the branch. This is because an increased flow of Solim?es water into the branch might lead to the branch's widening and becoming the main stream. This could have substantial environmental effects.