SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
robkacz
Topic Author
Posts: 2
Joined: September 27th, 2011, 6:29 pm

Multiple Regression with more than 16 variables

April 15th, 2014, 1:30 pm

I am trying to do a multiple regression with more than 16 variables and just realized that Microsfot Excel cannot handle the number of variables.I don't have access to any other software packages at the moment.Is anyone aware of any free packages that can handle this number of variables? Thanks
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Multiple Regression with more than 16 variables

April 15th, 2014, 2:10 pm

R
 
User avatar
riskneutralprob
Posts: 37
Joined: December 16th, 2005, 5:16 pm

Multiple Regression with more than 16 variables

April 15th, 2014, 2:15 pm

R would be my first choice.If you have access to matrix operations (ie inverse, etc), you could do it yourself. Maybe you could even use excel's MINVERSE method.
 
User avatar
acastaldo
Posts: 1416
Joined: October 11th, 2002, 11:24 pm

Multiple Regression with more than 16 variables

April 15th, 2014, 5:51 pm

 
User avatar
tagoma
Posts: 18379
Joined: February 21st, 2010, 12:58 pm

Multiple Regression with more than 16 variables

April 15th, 2014, 8:57 pm

QuoteOriginally posted by: robkaczI am trying to do a multiple regression with more than 16 variables multicollinearity what?
 
User avatar
dweeb
Posts: 142
Joined: July 11th, 2009, 8:10 pm

Multiple Regression with more than 16 variables

April 15th, 2014, 9:26 pm

parsimonious
 
User avatar
staassis
Posts: 30
Joined: April 12th, 2014, 5:10 pm

Multiple Regression with more than 16 variables

April 21st, 2014, 1:38 pm

Yes, R would be the best time investment for you even if you had already paid tons of money for packages like SAS.If you are ever interested in exploiting simple and complex data mining techniques for trading or something else, R has the richest and most up-to-date libraries.
 
User avatar
ExSan
Posts: 4583
Joined: April 12th, 2003, 10:40 am

Multiple Regression with more than 16 variables

May 7th, 2014, 11:19 am

what is the algorithm you are trying to implement ?
 
User avatar
bearish
Posts: 5800
Joined: February 3rd, 2011, 2:19 pm

Multiple Regression with more than 16 variables

May 7th, 2014, 11:02 pm

QuoteOriginally posted by: ExSanwhat is the algorithm you are trying to implement ?{=MMULT(MINVERSE(MMULT(TRANSPOSE(X),X)),MMULT(TRANSPOSE(X),Y))}
 
User avatar
ronnotel
Posts: 19
Joined: June 13th, 2006, 5:23 pm

Multiple Regression with more than 16 variables

May 8th, 2014, 2:55 am

A very nice open source implementation is available for free from alglib. This would be useful if you wanted something built into a code framework. Lots of control, ability to tune, etc.
 
User avatar
dweeb
Posts: 142
Joined: July 11th, 2009, 8:10 pm

Multiple Regression with more than 16 variables

May 8th, 2014, 10:57 am

QuoteOriginally posted by: bearishQuoteOriginally posted by: ExSanwhat is the algorithm you are trying to implement ?{=MMULT(MINVERSE(MMULT(TRANSPOSE(X),X)),MMULT(TRANSPOSE(X),Y))}It's missing a bracket.
 
User avatar
bearish
Posts: 5800
Joined: February 3rd, 2011, 2:19 pm

Multiple Regression with more than 16 variables

May 8th, 2014, 3:52 pm

QuoteOriginally posted by: dweebQuoteOriginally posted by: bearishQuoteOriginally posted by: ExSanwhat is the algorithm you are trying to implement ?{=MMULT(MINVERSE(MMULT(TRANSPOSE(X),X)),MMULT(TRANSPOSE(X),Y))}It's missing a bracket.Why don't you try it? I left the curly brackets on the outside to indicate that you need to hit Ctrl-Shift-Enter to invoke it. You actually have to remove them manually.
 
User avatar
dweeb
Posts: 142
Joined: July 11th, 2009, 8:10 pm

Multiple Regression with more than 16 variables

May 8th, 2014, 7:35 pm

QuoteOriginally posted by: bearishQuoteOriginally posted by: dweebQuoteOriginally posted by: bearishQuoteOriginally posted by: ExSanwhat is the algorithm you are trying to implement ?{=MMULT(MINVERSE(MMULT(TRANSPOSE(X),X)),MMULT(TRANSPOSE(X),Y))}It's missing a bracket.Why don't you try it? I left the curly brackets on the outside to indicate that you need to hit Ctrl-Shift-Enter to invoke it. You actually have to remove them manually.It was a joke. This is an Excel array for a a point estimate from a multiple explanatory variable regression??
 
User avatar
katastrofa
Posts: 9581
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Multiple Regression with more than 16 variables

May 10th, 2014, 6:07 pm

Evil/crap solution: if Excel cannot handle all 16 variables, you can fit 8 variables, calculate residuals and fit the residuals with the remaining 8 variables. The norm you will be minimising is different than in the original problem so you will get a different result, but hopefully similar to the right one :-)a
 
User avatar
DavidJN
Posts: 1757
Joined: July 14th, 2002, 3:00 am

Multiple Regression with more than 16 variables

May 12th, 2014, 3:35 pm

I once attended an academic job interview for which the candidate presented a paper that smelled a bit fishy, so to speak. A few probing questions revealed that the guy had almost as many regression variables as data points. Needless to say he didn't get the job!
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On