missing data in return series...

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

missing data in return series...

ShyhWeir Tzang
Dear all:

I have a portfolio of about 50 stocks of which about 10~15 stocks with
unequal lengths. That means they have shorter historical return series than
others. How may I estimate the covariance matrix and mean of the stocks? Is
the Stambaugh (1997) ("Analyzing investments whose histories differ in
length") method still valid for individual stocks instead of funds? Is there
any better way or more efficient way to estimate their mean and covariance
matrix? Any help or suggestion is highly appreciated.

--

        [[alternative HTML version deleted]]

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.
Reply | Threaded
Open this post in threaded view
|

Re: missing data in return series...

Patrick Burns-2
There are two functions in the BurStFin
package ('factor.model.stat' and 'var.shrink.eqcor')
that will create variance matrix estimates
when there are missing values in the return
matrix.

The second of those gives Ledoit-Wolf estimates,
and is probably going to give the more useful
results.

I believe that the best way to handle missing values
for these estimates is still an open research question.
The functions handle missing values, no claim that
they do it optimally.

You can get the package via:

install.packages('BurStFin', repos='http://www.burns-stat.com/R')

As for means: historical means are in general not
of much use, so it is unlikely that it will matter
how you estimate them.


On 19/09/2011 08:26, ShyhWeir Tzang wrote:

> Dear all:
>
> I have a portfolio of about 50 stocks of which about 10~15 stocks with
> unequal lengths. That means they have shorter historical return series than
> others. How may I estimate the covariance matrix and mean of the stocks? Is
> the Stambaugh (1997) ("Analyzing investments whose histories differ in
> length") method still valid for individual stocks instead of funds? Is there
> any better way or more efficient way to estimate their mean and covariance
> matrix? Any help or suggestion is highly appreciated.
>
> --
>
> [[alternative HTML version deleted]]
>
> _______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> -- Subscriber-posting only. If you want to post, subscribe first.
> -- Also note that this is not the r-help list where general R questions should go.
>

--
Patrick Burns
[hidden email]
http://www.burns-stat.com
http://www.portfolioprobe.com/blog
twitter: @portfolioprobe

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.
Reply | Threaded
Open this post in threaded view
|

Re: missing data in return series...

Eric Zivot
In reply to this post by ShyhWeir Tzang
I would look at the package monomvn: Estimation for multivariate normal and
Student-t data with monotone missingness. Professor Gramacy gave a very nice
presentation at the recent R/Finance 2011 conference:
http://www.rinfinance.com/agenda/2011/RobertGramacy.pdf



-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of ShyhWeir Tzang
Sent: Monday, September 19, 2011 12:27 AM
To: [hidden email]
Subject: [R-SIG-Finance] missing data in return series...

Dear all:

I have a portfolio of about 50 stocks of which about 10~15 stocks with
unequal lengths. That means they have shorter historical return series than
others. How may I estimate the covariance matrix and mean of the stocks? Is
the Stambaugh (1997) ("Analyzing investments whose histories differ in
length") method still valid for individual stocks instead of funds? Is there
any better way or more efficient way to estimate their mean and covariance
matrix? Any help or suggestion is highly appreciated.

--

        [[alternative HTML version deleted]]

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions
should go.

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.
Reply | Threaded
Open this post in threaded view
|

Re: missing data in return series...

ShyhWeir Tzang
I will look into this issue and many thanks to Eric for your great advice...

2011/9/20 Eric Zivot <[hidden email]>

> I would look at the package monomvn: Estimation for multivariate normal and
> Student-t data with monotone missingness. Professor Gramacy gave a very
> nice
> presentation at the recent R/Finance 2011 conference:
> http://www.rinfinance.com/agenda/2011/RobertGramacy.pdf
>
>
>
> -----Original Message-----
> From: [hidden email]
> [mailto:[hidden email]] On Behalf Of ShyhWeir Tzang
> Sent: Monday, September 19, 2011 12:27 AM
> To: [hidden email]
> Subject: [R-SIG-Finance] missing data in return series...
>
> Dear all:
>
> I have a portfolio of about 50 stocks of which about 10~15 stocks with
> unequal lengths. That means they have shorter historical return series than
> others. How may I estimate the covariance matrix and mean of the stocks? Is
> the Stambaugh (1997) ("Analyzing investments whose histories differ in
> length") method still valid for individual stocks instead of funds? Is
> there
> any better way or more efficient way to estimate their mean and covariance
> matrix? Any help or suggestion is highly appreciated.
>
> --
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> -- Subscriber-posting only. If you want to post, subscribe first.
> -- Also note that this is not the r-help list where general R questions
> should go.
>
>

--
Shyh-Weir Tzang
Finance Department
Asia University
Tel: 04-2332-3456#48055
Mobile: 0929125845/0973830823

»N¥Kºû
¨È¬w¤j¾Ç°]°Èª÷¿Ä¨t
¹q¸Ü¡G04-2332-3456#48055
¤â¾÷¡G0929125845/0973830823
41354¥x¤¤Ãú®p¬hÂ׸ô500¸¹

        [[alternative HTML version deleted]]


_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.
Reply | Threaded
Open this post in threaded view
|

Re: missing data in return series...

ShyhWeir Tzang
In reply to this post by Patrick Burns-2
Thanks to Patrick for your great suggestions and I will try the package to
see if it really works for me.

2011/9/19 Patrick Burns <[hidden email]>

> There are two functions in the BurStFin
> package ('factor.model.stat' and 'var.shrink.eqcor')
> that will create variance matrix estimates
> when there are missing values in the return
> matrix.
>
> The second of those gives Ledoit-Wolf estimates,
> and is probably going to give the more useful
> results.
>
> I believe that the best way to handle missing values
> for these estimates is still an open research question.
> The functions handle missing values, no claim that
> they do it optimally.
>
> You can get the package via:
>
> install.packages('BurStFin', repos='http://www.burns-stat.**com/R<http://www.burns-stat.com/R>
> ')
>
> As for means: historical means are in general not
> of much use, so it is unlikely that it will matter
> how you estimate them.
>
>
>
> On 19/09/2011 08:26, ShyhWeir Tzang wrote:
>
>> Dear all:
>>
>> I have a portfolio of about 50 stocks of which about 10~15 stocks with
>> unequal lengths. That means they have shorter historical return series
>> than
>> others. How may I estimate the covariance matrix and mean of the stocks?
>> Is
>> the Stambaugh (1997) ("Analyzing investments whose histories differ in
>> length") method still valid for individual stocks instead of funds? Is
>> there
>> any better way or more efficient way to estimate their mean and covariance
>> matrix? Any help or suggestion is highly appreciated.
>>
>> --
>>
>>        [[alternative HTML version deleted]]
>>
>> ______________________________**_________________
>> [hidden email] mailing list
>> https://stat.ethz.ch/mailman/**listinfo/r-sig-finance<https://stat.ethz.ch/mailman/listinfo/r-sig-finance>
>> -- Subscriber-posting only. If you want to post, subscribe first.
>> -- Also note that this is not the r-help list where general R questions
>> should go.
>>
>>
> --
> Patrick Burns
> [hidden email]
> http://www.burns-stat.com
> http://www.portfolioprobe.com/**blog <http://www.portfolioprobe.com/blog>
> twitter: @portfolioprobe
>


--
Shyh-Weir Tzang
Finance Department
Asia University
Tel: 04-2332-3456#48055
Mobile: 0929125845/0973830823

»N¥Kºû
¨È¬w¤j¾Ç°]°Èª÷¿Ä¨t
¹q¸Ü¡G04-2332-3456#48055
¤â¾÷¡G0929125845/0973830823
41354¥x¤¤Ãú®p¬hÂ׸ô500¸¹

        [[alternative HTML version deleted]]


_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.