This is a common question I usually face while processing historical data. I have year on year data of an event for the past N years.I would like to assign weights to the data of these N years so that the data corresponding to the most recent year as the highest weight and the data corresponding to the oldest year has the least. This is to give more importance to recent trend as compared to very old trends.
Questions
Is there any scientific way to assign weights to the years i.e. what should be the weight for the most recent year, what should be the weight for the previous year an so on?
Is there any scientific justification in assigning equal weights to the preceding historical data after a certain point?
Is there any scientific justification to ignore the preceding historical data after a certain point i.e. weight = 0?
I think the answers will depend on the exact nature of the data under study and we might not have a general answer that will work with all data. Even then can we at least have a rule of thumb which is independent of the actual data. Any reference or pointers to literature would also be helpful.