|
PureBytes Links
Trading Reference Links
|
Andrew,
A very effective detrend technique is to simply take the first differences
of the data then look at the resulting series and see if it "trends".
Typically, that's all that's needed. However, if trend still exists, take
second differences (first difference of the first difference). Note that
second differences can be used as a measure of momentum in stock prices.
Regards,
Bill Vedder
Andrew wrote:
> I would like to begin a new discussion on this list regarding the
> detrending of a price series.
>
> I would imagine that the following methods would do a fair job of
> removing the linear trend component from a time series.
>
> 1. Subtract series average from each data point.
> 2. Use a bandpass filter to filter out all information above and below
> certain frequency cut off points.
> 3. Calculate a least squares regression line of the time series and
> subtract
> it from each data point
>
> I would imagine that the optimal method would be the lease squares
> regression line. Would anyone here on the list comment on the above?
> What is the best way for "detrending" data in your opinion. I am
> looking forward to an educational discussion from some of the more
> mathematically inclined members from the list.
>
> Regards,
>
> Andrew
|