Basic OLS question?
in [Matlab]
|
From: Oleg Komarov on 15 Feb 2010 18:50 "Antony zerrar" <zerrar(a)yahoo.com> wrote in message <hlc9om$618$1(a)fred.mathworks.com>... > "Oleg Komarov" <oleg.komarovRemove.this(a)hotmail.it> wrote in message <hlbu1m$5ei$1(a)fred.mathworks.com>... > > "Antony zerrar" <zerrar(a)yahoo.com> wrote in message <hlbqdf$bkb$1(a)fred.mathworks.com>... > > > Hi all, > > > > > > There are different approaches in estimating OLS in matlab, and would like to ask to ask which approach to use to derive t-stats or sd errors of the coefficients? > > > > > > Thank you. > > > > If you have the statistic toolbox: > > - help regstats > > - http://www.mathworks.com/matlabcentral/fileexchange/26169-regstats2 (also needs stat TB) > > > > Oleg > > Thanks for the replies, I am having a problem with regstats as the coefficients given and their t-stats values are not correct (compared to paper solution, or even other programs or even within the regress command). any ideas? You have to note that while regress asks explicitly to add a column of ones for the intercept, regstat does this by itself. So remove the column of ones if you're using regstats (or wait till my fcn - regstats2 - will be updated with the new 'onlydata' option). Oleg
From: Antony zerrar on 23 Feb 2010 04:48 "Oleg Komarov" <oleg.komarovRemove.this(a)hotmail.it> wrote in message <hlcmjs$bps$1(a)fred.mathworks.com>... > "Antony zerrar" <zerrar(a)yahoo.com> wrote in message <hlc9om$618$1(a)fred.mathworks.com>... > > "Oleg Komarov" <oleg.komarovRemove.this(a)hotmail.it> wrote in message <hlbu1m$5ei$1(a)fred.mathworks.com>... > > > "Antony zerrar" <zerrar(a)yahoo.com> wrote in message <hlbqdf$bkb$1(a)fred.mathworks.com>... > > > > Hi all, > > > > > > > > There are different approaches in estimating OLS in matlab, and would like to ask to ask which approach to use to derive t-stats or sd errors of the coefficients? > > > > > > > > Thank you. > > > > > > If you have the statistic toolbox: > > > - help regstats > > > - http://www.mathworks.com/matlabcentral/fileexchange/26169-regstats2 (also needs stat TB) > > > > > > Oleg > > > > Thanks for the replies, I am having a problem with regstats as the coefficients given and their t-stats values are not correct (compared to paper solution, or even other programs or even within the regress command). any ideas? > > You have to note that while regress asks explicitly to add a column of ones for the intercept, regstat does this by itself. So remove the column of ones if you're using regstats (or wait till my fcn - regstats2 - will be updated with the new 'onlydata' option). > > Oleg Thanks for the replies, apart that I dont require a constant, did tried it but still the coefficients and tstats are way off :( Will wait for regstats2 be complete
From: Oleg Komarov on 23 Feb 2010 05:01 regstats2 is complete in the sense that yuo can now supply just the data without adding the constant: http://www.mathworks.com/matlabcentral/fileexchange/26169-regstats2 (also needs stat TB) example: OUT = regstats2(y,x,'onlydata',{'tstat','beta'}); Oleg
First
|
Prev
|
Pages: 1 2 Prev: Compiled Demo of MatLab Graphics? Next: mex file define a sparse matrix |