[R] multivariate multiple regression coefficient

[서은경]

   hi,

   I considered multivariate multiple regression for respons, predictor, each 5
   dimension.

   on going process, through


   initb <- eigen(temp)$vectors[,1:3]
   temp <- vhalf%*%Kproduct(Ir,initb)
   (herein 'vhalf' is root for inverse of covariance matrix

   and 'Kproduct' is Kronecker product)


   and then process regression. However,

   I have tried using the 'lm' function and i get the result, like this. ( 'xi'
   is 25 by 1 vector )


   > reg <- lm(xi~temp)
   >
   > reg

   Call:
   lm(formula = xi ~ temp)

   Coefficients:
   (Intercept)        temp1        temp2        temp3        temp4
   temp5        temp6        temp7
    -3.704e-15    2.434e-02    4.087e-02           NA   -5.241e-01
   -1.213e-02           NA   -1.960e-02
         temp8        temp9       temp10       temp11       temp12
   temp13       temp14       temp15
     6.516e-03           NA   -6.651e-02    8.865e-02           NA
   4.272e-02   -3.085e-02           NA



   Coefficients for multiple of three are 'NA'.

   I'd like get any value, so i directly tried to calculate using matrix.

   That result,


   > ginv(t(temp)%*%temp)%*%t(temp)%*%xi
                  [,1]
    [1,]  0.0239396255
    [2,]  0.0402622882
    [3,] -0.0058317229
    [4,] -0.5216032391
    [5,] -0.0083355651
    [6,]  0.0362961561
    [7,] -0.0195581411
    [8,]  0.0065824700
    [9,]  0.0006407087
   [10,] -0.0668295201
   [11,]  0.0881623108
   [12,] -0.0046706822
   [13,]  0.0427419205
   [14,] -0.0308143260
   [15,]  0.0003160107



   why don't I get coefficient value using the 'lm' function?

   and what's different to two way?


   I really need your help.

   I waiting your answer !







   [1][Nmn9OQFrSBLEKLcn6umQCQ00]
   [@from=sluvsek&rcpt=r%2Dhelp%40r%2Dproject%2Eorg%2E&msgid=%3C20100924155914%
   2EHM%2EM000000000E5rHm%40sluvsek%2Ewwl1049%2Ehanmail%2Enet%3E]

References

   1. mailto:sluvsek at hanmail.net