Re: time varying correlation
- From: Richard Ulrich <Rich.Ulrich@xxxxxxxxxxx>
- Date: Tue, 21 Jun 2005 14:10:48 -0400
On 21 Jun 2005 03:55:23 -0700, "sharad gupta"
<sharad_gupta90@xxxxxxxxxxx> wrote:
> Hi!
>
> I am trying to find out relation b/w two variables, possibly look at
> the correlation and see if the relationship has been strong in the past
> and recently it is broken... is there any statistical method to figure
> this out? I guess a crude way of doing this is look at the long term
> (say 1 year data) and short term (day 1 month) correlation...
You are looking at time series. Time series are a great source
for spurious correlations. One of the first questions should be,
"How big are the autocorrelations within each series?"
It sounds like you could get a quick overview of whatever
is happening with two series by plotting them each against time.
Does either one look strange?
If that doesn't tell you what is going on, you might learn
something by looking at cross-correlation r's for each the
12 months separately.
>
> Further :
> It is daily data - I have around 1 year data available...so if I try
> with monthly or weekly data, there will be very few points and that may
> not give a good estimate of correlation. Will moving average be better?
>
> What is the statistical technique to analize this kind of data( I would
> guess temporal correlation, but don't how to calculate it).
--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.
- References:
- time varying correlation
- From: sharad gupta
- time varying correlation
- Prev by Date: Re: Finding 2D probability distribution
- Next by Date: Re: Finding 2D probability distribution
- Previous by thread: Re: time varying correlation
- Next by thread: Re: time varying correlation
- Index(es):
Relevant Pages
|
