When we analyze the multiple time series of which variables are related with each other by constraints, it is necessary to analyze those variables in a whole. The reason is that if we calculate the estimates of each time series independently, it happens that the result does not satisfy the given constraints.
For example, in the time series of market shares, the variables are related with each other by one constraint that the sum of all concurrent proportions is one. Korean automobile market is divided into three segments, those are small-sized, middle-sized, and large-sized car, and the proportion of each segment changes slowly as time goes by. This paper proposes a Bayesian method to detect structural changes in market share time series in domestic automobile market, especially for level shift and drift shift, and proves its exellence by comparing it with other approaches.