QuantLib › quantlib-users

Sequence Statistics

Classic

List

Threaded

6 messages Options

simone pilozzi

Sequence Statistics

Dear Users,

I was wondering if there's a rapid method to update a sequence statistic.

The purpose is to compute a covariance matrix from a dataset of time series using a "rolling window" with constant length and floating start/end date.

It would be then helpful to start with an initial dataset of n time series and, at each step, add the new observation and drop the last ,instead of recreate the sequence statistics object from scratch.

Is it feasible? Or does it exist another "smart" solution?

Thanks in advance

Simone

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense..
http://p.sf.net/sfu/splunk-d2d-c1
_______________________________________________
QuantLib-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-users

Circo Giuseppe (DAM)

Re: Sequence Statistics

Hi,

you can give a binary weight of 0 or 1 to your datas that will be updated with your rolling period,

use qlSequenceStatistics2

regards,

From: simone pilozzi [mailto:[hidden email]]
Sent: 24 June 2011 09:44
To: [hidden email]; [hidden email]
Subject: [Quantlib-users] Sequence Statistics

Dear Users,

I was wondering if there's a rapid method to update a sequence statistic.

The purpose is to compute a covariance matrix from a dataset of time series using a "rolling window" with constant length and floating start/end date.

Is it feasible? Or does it exist another "smart" solution?

Thanks in advance

Simone

simone pilozzi

Re: Sequence Statistics

Thanks,

But is a QuantLibXL object isn't it ? (my routine is C++).
Moreover, is there a method to change only the weights and not the sample data?

Thanks again

On 24 June 2011 09:54, Circo Giuseppe (DAM) <[hidden email]> wrote:

Hi,

you can give a binary weight of 0 or 1 to your datas that will be updated with your rolling period,

use qlSequenceStatistics2

regards,

From: simone pilozzi [mailto:[hidden email]]
Sent: 24 June 2011 09:44
To: [hidden email]; [hidden email]
Subject: [Quantlib-users] Sequence Statistics

Dear Users,

I was wondering if there's a rapid method to update a sequence statistic.

The purpose is to compute a covariance matrix from a dataset of time series using a "rolling window" with constant length and floating start/end date.

It would be then helpful to start with an initial dataset of n time series and, at each step, add the new observation and drop the last ,instead of recreate the sequence statistics object from scratch.

Is it feasible? Or does it exist another "smart" solution?

Thanks in advance

Simone

simone pilozzi

Re: Sequence Statistics

Thanks,

I did not find a specific method to change only the weights of the sequence...

On 24 June 2011 13:46, Circo Giuseppe (DAM) <[hidden email]> wrote:

yes indeed i was meaning in the xl version,

my idea is that you upload (time serie) all the matrix 1 time, and that you give a weight to each item in the time serie of 0 or 1 depending wich period you want to analyse,

with a if date btw x and y then weight = 1 else weight =0

very simple idea and i guess the function should also exist in c++

cheers,

From: simone pilozzi [mailto:[hidden email]]
Sent: 24 June 2011 13:42
To: Circo Giuseppe (DAM)
Cc: [hidden email]; [hidden email]
Subject: Re: [Quantlib-users] Sequence Statistics

Thanks,

But is a QuantLibXL object isn't it ? (my routine is C++).
Moreover, is there a method to change only the weights and not the sample data?

Thanks again

On 24 June 2011 09:54, Circo Giuseppe (DAM) <[hidden email]> wrote:

Hi,

you can give a binary weight of 0 or 1 to your datas that will be updated with your rolling period,

use qlSequenceStatistics2

regards,

From: simone pilozzi [mailto:[hidden email]]
Sent: 24 June 2011 09:44
To: [hidden email]; [hidden email]
Subject: [Quantlib-users] Sequence Statistics

Dear Users,

I was wondering if there's a rapid method to update a sequence statistic.

The purpose is to compute a covariance matrix from a dataset of time series using a "rolling window" with constant length and floating start/end date.

It would be then helpful to start with an initial dataset of n time series and, at each step, add the new observation and drop the last ,instead of recreate the sequence statistics object from scratch.

Is it feasible? Or does it exist another "smart" solution?

Thanks in advance

Simone

Luigi Ballabio

Re: Sequence Statistics

In reply to this post by simone pilozzi

On Fri, 2011-06-24 at 09:44 +0200, simone pilozzi wrote:
> I was wondering if there's a rapid method to update a sequence
> statistic.
> The purpose is to compute a covariance matrix from a dataset of time
> series using a "rolling window" with constant length and floating
> start/end date.
> It would be then helpful to start with an initial dataset of n time
> series and, at each step, add the new observation and drop the
> last ,instead of recreate the sequence statistics object from scratch.
> Is it feasible? Or does it exist another "smart" solution?

No, sorry. SequenceStatistics doesn't provide the means to remove a
datum or update its weight, so it can't be used this way. It does make
sense for your use case to have a statistics class working the way you
describe, but you'll need either to modify the existing class or write a
new one.

Luigi

--

I hate quotations.
-- Ralph Waldo Emerson

------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security
threats, fraudulent activity, and more. Splunk takes this data and makes
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2d-c2
_______________________________________________
QuantLib-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-users

simone pilozzi

Re: Sequence Statistics

Thanks Luigi

On 27 June 2011 16:58, Luigi Ballabio <[hidden email]> wrote:

On Fri, 2011-06-24 at 09:44 +0200, simone pilozzi wrote:
> I was wondering if there's a rapid method to update a sequence
> statistic.
> The purpose is to compute a covariance matrix from a dataset of time
> series using a "rolling window" with constant length and floating
> start/end date.
> It would be then helpful to start with an initial dataset of n time
> series and, at each step, add the new observation and drop the
> last ,instead of recreate the sequence statistics object from scratch.
> Is it feasible? Or does it exist another "smart" solution?

No, sorry. SequenceStatistics doesn't provide the means to remove a
datum or update its weight, so it can't be used this way. It does make
sense for your use case to have a statistics class working the way you
describe, but you'll need either to modify the existing class or write a
new one.

Luigi

--

I hate quotations.
-- Ralph Waldo Emerson

------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security
threats, fraudulent activity, and more. Splunk takes this data and makes
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2d-c2
_______________________________________________
QuantLib-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-users