[R] Help with predict.Arima with external regressor values [Repalced]

[Prof Brian Ripley]

First, xreg is documented as a vector or matrix, not a data frame.

Second, there may be a scoping problem, so try a more informative name 
than `y' when you do the fit.

Third, R has debugging facilities, so please use them to find out 
precisely what is wrong.

On Thu, 18 Dec 2003, Yves wrote:

> 
> Hi all there
> 
> I am enjoying R since 2 weeks and I come to my first deadlock, il am trying
> to use predict.Arima in the ts package.
> I get a "Error in cbind(...) : cannot create a matrix from these types"
> 
> -- Start R session -----------------------------------------------------
> 
> > fitdiv <- arima(data, c(2, 0, 3), xreg = y ) ; print(fitdiv)
> 
> Call:
> arima(x = data, order = c(2, 0, 3), xreg = y)
> 
> Coefficients:
>          ar1     ar2     ma1      ma2      ma3  intercept   EUSA1  EUSA10
> EUSA15  EUSA20    EUSA5  H15T10Y   H15T1Y  H15T20Y  H15T3M   H15T5Y   USSW10
>       -0.001  0.6502  0.3328  -0.5021  -0.1135    -0.0535  0.0469
> 0.0075  -0.0263  0.0299  -0.0344   0.1012  -0.0382   0.0092
> 0.0385  -0.0757  -0.1577
> s.e.   0.523  0.4002  0.5262   0.2711   0.0828     0.1027  0.0308  0.0802
> 0.0931  0.0743   0.0414   0.0469   0.0215   0.0360  0.0276   0.0344   0.0477
>       USSW15   USSW20  USSW30   USSW5  CAC.INDEX  DAX.INDEX  MIB30.INDEX
> OMX.INDEX  SX5P.INDEX  UKX.INDEX  VDAX.INDEX  VIX.INDEX
>       0.0254  -0.0141  0.0133  0.1186    -0.1816     0.0652
>    0.0848    -0.1836      0.1134    -0.1742      0.0236    -0.0482
> s.e.  0.0588   0.0251  0.0363  0.0278     0.0907     0.0528       0.0860
> 0.0516      0.1518     0.1025      0.0591     0.0470
> 
> sigma^2 estimated as 1.258:  log likelihood = -762.3,  aic = 1584.59
> >
> > fordiv <- predict(fitdiv, n.ahead = 1, newxreg = newregy ,  se.fit = TRUE)
> Error in cbind(...) : cannot create a matrix from these types
> >
> >
> > str(data)
>  num [1:497] -0.34 -1.36 -0.5 -0.46 0.01 0.1 0.68 0.06 0.16 0.48 ...
> >
> > str(newregy)
>  num [1, 1:23] -0.6 -0.3 0.15 1.08 -1.8 3 2 3 0 5 ...
>  - attr(*, "dimnames")=List of 2
>   ..$ : chr "498"
>   ..$ : chr [1:23] "EUSA1" "EUSA10" "EUSA15" "EUSA20" ...
> >
> > str(y)
> `data.frame':   497 obs. of  23 variables:
>  $ EUSA1      : num  0.7 5.9 -0.6 1.8 5.7 1.9 0.5 -6.6 2.5 2.3 ...
>  $ EUSA10     : num  -4.5 3.8 -11.7 3.2 4.2 -5.4 -2.2 -6.5 0.8 2 ...
>  $ EUSA15     : num  -5.4 3.6 -11 3.7 3.4 -4.3 -3.9 -4.7 0.3 2.6 ...
>  $ EUSA20     : num  -5 3.6 -10.8 4.3 2.3 -4.1 -3.5 -5 0 3.1 ...
>  $ EUSA5      : num  -4.3 5.4 -10.8 2.5 6.3 -2.4 -1.3 -6.6 2.3 -0.2 ...
>  $ H15T10Y    : num  -4 2 -9 1 0 -10 -8 -1 -3 0 ...
>  $ H15T1Y     : num  -4 1 -6 0 -2 -7 -12 2 -1 2 ...
>  $ H15T20Y    : num  -3 4 -11 1 -1 -12 -4 2 -6 2 ...
>  $ H15T3M     : num  -1 -1 -4 0 0 0 -10 0 2 1 ...
>  $ H15T5Y     : num  -4 2 -11 0 -1 -11 -13 1 -1 2 ...
>  $ USSW10     : num  -8.6 0.6 -10.5 2.4 -2.9 -5.8 -15 0.4 -3.5 1 ...
>  $ USSW15     : num  -7.9 1.1 -9.8 2 -3.1 -5.8 -13.2 2 -4.4 2.1 ...
>  $ USSW20     : num  -6.7 1.4 -9.5 1.3 -3 -6 -11 1.2 -4.2 2.9 ...
>  $ USSW30     : num  -6.2 1.3 -8.4 1.6 -2.6 -6.9 -8.5 -0.9 -2.8 2.2 ...
>  $ USSW5      : num  -7.7 -0.4 -12.7 0.9 -4 -7.4 -17.5 0.9 0.3 1.6 ...
>  $ CAC.INDEX  : num  2.18 0.03 -1.45 -1.03 0.41 -1.57 0.86 -2.24 1.44 -2.08
> ...
>  $ DAX.INDEX  : num  1.96 0.91 -1.64 0.08 0.99 -1.14 -0.35 -2.81 -0.08 -1.55
> ...
>  $ MIB30.INDEX: num  2.08 -0.48 -0.94 0.82 0.22 -2.22 0.8 -2.5 1.34 -1.35
> ...
>  $ OMX.INDEX  : num  4.07 0.28 -0.56 -2.47 -0.16 -1.58 1.13 -3.5 -1.15 -1.85
> ...
>  $ SX5P.INDEX : num  1.95 0.1 -1.22 -1.01 -0.31 -0.96 0.84 -2.71 1.18 -1.05
> ...
>  $ UKX.INDEX  : num  1.91 0.09 -0.57 -0.82 -0.42 -0.73 0.15 -1.65 1.02 -0.75
> ...
>  $ VDAX.INDEX : num  -1.59 -0.74 0.73 -0.36 -0.36 0.87 -0.51 1.64 -0.02 0.85
> ...
>  $ VIX.INDEX  : num  -1.37 -0.89 1.22 0.16 0.3 -0.1 0.57 0.98 -0.88 0.75 ...
> >
> > fordiv <- predict(fitdiv, n.ahead = 1, newxreg = matrix(0,1,23) ,  se.fit
> = TRUE)
> Error in cbind(...) : cannot create a matrix from these types
> >
> 
> -- End R session -----------------------------------------------------
> 
> I also tried to replace newregy by a matrix of zeros  matrix(0,1,23)
> 
> Please tell if I am doing something wrong ... I did not see any example in
> the help about external regressor so I had to start from scratch.
> 
> Best regards
> 
> Yves Oloui.
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> 
> 

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595