Skip to content

correlation matrix

7 messages · Patrick Burns, debashis dutta, Enrico Schumann +1 more

#
Can you be more specific about what the
problem is, and perhaps tell us what
you've tried that is unsatisfactory?
On 27/11/2011 09:06, debashis dutta wrote:

  
    
#
Hi Debashis, I dont think there is any direct implementation in R for that.
What you need is to have a PD matrix for the underlying VCV matrix, which is
not the case for yours. Do you have lot of missing data? May be you can try
with constructing VCV matrix after considering pairwise variables and then
tweak the eigen value little bit to make your estimated VCV matrix PD.

One algorithm for that may be like (if your estimated matrix is NND):

Mod_VCV = Original_VCV + (Identity_Mat(n) - Original_VCV)*(Lambda/(1-Lambda)
+ a_very_small_positive_number)

Lambda is the smallest eigen value for you estimated VCV matrix.

HTH,

Thanks and regards,
_____________________________________________________

Arun Kumar Saha, FRM
QUANTITATIVE RISK AND HEDGE CONSULTING SPECIALIST
Visit me at: http://in.linkedin.com/in/ArunFRM
_____________________________________________________

--
View this message in context: http://r.789695.n4.nabble.com/correlation-matrix-tp4112075p4112158.html
Sent from the Rmetrics mailing list archive at Nabble.com.
#
Am 27.11.2011 10:52, schrieb Arun.stat:
Just for the record: see for example function 'nearPD' in package 
Matrix, or 'repairMatrix' in package NMOF.

  
    
#
Dear Enrico & Arun,

Thanks for the help. I will try for the same.

Best regards
Debashis
On 27/11/2011, Arun Kumar Saha <arun.kumar.saha at gmail.com> wrote: