Re: Time Series Observation Identification

[Jody McIntyre <jody_mcintyre99@yahoo.com>]

Dear S-News,

 

I have a very large time series dataset with several variables, for example:

 

Y₁           x₁            x₂ ??????????_.

_.

^Y_n            x_1n        x_2n

 

where x₁ ?..x_n  can be row number, year, month day etc.

 

I would like to write a routine (or perhaps 2 if necessary) in S-Plus (Windows Version 6.1) to identify the nth highest (lowest) maximum (minimum) y for a given time period, say year. If there are any missing values (i.e. less than 'n' non-missing values), the routine would need to be aware of this and function accordingly.

 

Having identified the above, I would then like to calculate the 5 year moving average. For example, if in the above routine I were to identify the 12^th highest value y for every year 1-23, I would then like to calculate a moving average of span 5 [(y_t, y_t-1, y_t-2, y_t-3, y_t-4))/ 5]. Clearly, some averages would not be able to be calculated (i.e. t<5) because there would not be enough observations available.

 

I have been told that it would be possible to carry out such manoeuvres in STATA using the syntax _n and _N and the 'sort' command but I wish to use S-Plus for its other time series capabilities.

 

Any insights gratefully acknowledged. I will publish any responses received.

 

Jody

---

 Yahoo! FareChase - Search multiple travel sites in one click.