Good evening everybody, I have a big problem and i will need our help. Please if anybody knows any idea to solve it: tell me!Well, let it be s1,…,s12, monthly coefficients that sums 1. This coefficients collection appears during 4 years and consenquently changing,ie coefficients from one year to another are different but the constrait remains in this year. My question is how can I modelling the behaviour of this time series with the condition: sum(coefficients in one year)=1. I’ve tried with ARIMA but the condition doesn’t remain besides the R^2 is very low about 50%. Do you know any method to solve this problem?Thanks.

one way would be use to a kalman filterapproach ( see andrew harvey's blue book,i forget title )and model the series using his "basic structural model" ( bsm ).this type of model imposes "sum constraints" sort ofautomatically.if you aren't familar with thekalman filter/dlms, it'sprobably not worthgetting involved with the kalmanfilter if it's just a homework assignment ?because it's a lot of stuff to digest.

First of all, a lot of thanks for your answer!! unfortunately this modelling isn't homework, so I'll have to start study kalman filters.If you could show me webs (pdf, docs) in which tackle this topic i'll thank for.Bye