The Math Forum

Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Math Forum » Discussions » sci.math.* » sci.stat.math

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: Calculating a smooth 90% limit for differences in a time series
Replies: 1   Last Post: Mar 1, 2014 12:57 PM

Advanced Search

Back to Topic List Back to Topic List Jump to Tree View Jump to Tree View   Messages: [ Previous | Next ] Topics: [ Previous | Next ]
Richard Ulrich

Posts: 2,961
Registered: 12/13/04
Re: Calculating a smooth 90% limit for differences in a time series
Posted: Mar 1, 2014 12:57 PM
  Click to see the message monospaced in plain text Plain Text   Click to reply to this topic Reply

On Fri, 28 Feb 2014 05:38:43 -0800 (PST),

>I have 50 data sets. Each set has three related time series: fast, medium, slow. My end purpose is simple, I want to generate a number that indicates a relative degree of change of the time series at each point. That relative degree of change should range between 0-1 for all the time series and all the data sets. The scales of the data set range from .0001 to 100.

I can't say that I understand your problem well.

However - given that the scales range fro 0.0001 to 100,
and that you want "relative degree of change", the very
first starting point would seem to be: Take the logarithm
of everything.

The standard deviation of scores, for log scaling, is
often used in various contexts as the same thing as
a coefficient of variation. It looks like you might have
something useful if you look at the SD of the change

- Are those consistent for different parts of a single
time series? If they change, you might not have
data consistent with the sort of model you are trying

>To accomplish this, I calculate the differences in a time series, delta(ts)=ts(t) - ts(t-1). Now I am trying to calculate an upper limit of 90% of those deltas. In other words, draw a smooth line on those differences such that only about 10% of the differences exceed that line. I will use that 90% limit to establish a maximum to normalize the differences between 0-1. Is this the best way to do this?
>I've been working on this for months, mostly linear programmatic methods, with no success. And trying to get it to work across 50 data sets is killing me. I'm sure there has to be an elegant mathematical way to do this. I can't be the first guy in town trying to normalize a relative degree of change of a time series.
>Any help or directions for research are greatly appreciated! Obviously my math skills are weak so examples would be most helpful. Thank you everyone for your time and brain power!

Econometricians are the folks who I know of, who are
most concerned with time series and their changes.
You might look for that sort of reference, or that
sort of List for posting the question.

Rich Ulrich

Point your RSS reader here for a feed of the latest messages in this topic.

[Privacy Policy] [Terms of Use]

© The Math Forum at NCTM 1994-2018. All Rights Reserved.