[R-SIG-Finance] Vector autoregression with Newey-West standard errors

[Liviu Andronic]

Dear all,

I'm currently fitting vector autoregression using VAR() from package
`vars'. It estimates VAR by using OLS, and by default it provides
"naive" standard errors (not HC and not HAC).

> require(vars)
> data(Canada)
> Canada <- (as.data.frame(Canada))
> head(Canada)
      e  prod    rw    U
1 929.6 405.4 386.1 7.53
2 929.8 404.6 388.1 7.70
3 930.3 403.8 390.5 7.47
4 931.4 404.2 394.0 7.27
5 932.7 405.0 396.8 7.37
6 933.6 404.4 400.0 7.13
> temp <- VAR(Canada[,1:3], p = 1, type = "const")
> coef(temp)
$e
        Estimate Std. Error t value  Pr(>|t|)
e.l1    0.923983   0.023492 39.3313 1.202e-53
prod.l1 0.155238   0.028725  5.4043 6.697e-07
rw.l1   0.008184   0.006824  1.1992 2.340e-01
const   5.255418  13.451377  0.3907 6.971e-01

$prod
        Estimate Std. Error t value  Pr(>|t|)
e.l1    -0.06662   0.032275  -2.064 4.229e-02
prod.l1  1.02613   0.039463  26.002 1.777e-40
rw.l1    0.02744   0.009376   2.926 4.478e-03
const   40.29543  18.480012   2.180 3.220e-02

$rw
        Estimate Std. Error t value  Pr(>|t|)
e.l1      0.1568    0.03650   4.296 4.907e-05
prod.l1  -0.1394    0.04463  -3.124 2.496e-03
rw.l1     0.9362    0.01060  88.303 1.012e-80
const   -62.0937   20.89779  -2.971 3.928e-03


However, I would like to obtain Heteroskedasticity-Autocorrelation
Consistent standard errors using NeweyWest() from package `sandwich',
which handles principally `lm' and `glm' objects. I noticed that the
VAR() returns an object of class `varest', which contains a list of
fitted `lm' objects. So I tried to apply NeweyWest() to individual
`lm' components of `varest', unsuccessfully.

> class(temp)
[1] "varest"
> class(temp$varresult$e)
[1] "lm"
> temp.lm <- temp$varresult$e
> class(temp.lm)
[1] "lm"
> require(sandwich)
> NeweyWest(temp.lm)
Error in AA %*% t(X) : requires numeric matrix/vector arguments
In addition: Warning message:
In ar.ols(x, aic = aic, order.max = order.max, na.action = na.action,  :
  model order: 1singularities in the computation of the projection
matrixresults are only valid up to model order0
> vcovHAC(temp.lm)
Error in bread. %*% meat. : non-conformable arguments


I would have expected the above to have worked. For "standard" `lm'
objects, I never had any such issues:
> temp.std.lm <- lm(e ~ prod + rw, data=Canada)
> class(temp.std.lm)
[1] "lm"
> NeweyWest(temp.std.lm)    # the variance-covariance matrix of the estimates
            (Intercept)      prod       rw
(Intercept)   30738.384 -83.52869  7.65428
prod            -83.529   0.23111 -0.02458
rw                7.654  -0.02458  0.00539
> require(lmtest)
> coeftest(temp.std.lm, df = Inf, vcov = NeweyWest)    # NW std errs for regression coefs

z test of coefficients:

            Estimate Std. Error z value Pr(>|z|)
(Intercept) 476.4344   175.3237    2.72  0.00658 **
prod          0.8819     0.4807    1.83  0.06659 .
rw            0.2454     0.0734    3.34  0.00083 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1


Could anyone suggest a way to work around this error? Thank you,
Liviu

PS I am also cc-ing this message to the authors of `vars' and
`sandwich', Bernhard Pfaff and Achim Zeileis, respectively. Perhaps
they could shed some light on the issue.





> sessionInfo()
R version 2.8.1 (2008-12-22)
i686-pc-linux-gnu

locale:
LC_CTYPE=en_GB.UTF-8;LC_NUMERIC=C;LC_TIME=en_GB.UTF-8;LC_COLLATE=en_GB.UTF-8;LC_MONETARY=C;LC_MESSAGES=en_GB.UTF-8;LC_PAPER=en_GB.UTF-8;LC_NAME=C;LC_ADDRESS=C;LC_TELEPHONE=C;LC_MEASUREMENT=en_GB.UTF-8;LC_IDENTIFICATION=C

attached base packages:
 [1] grid      splines   tcltk     stats     graphics  grDevices datasets
 [8] utils     methods   base

other attached packages:
 [1] Matrix_0.999375-23              relimp_1.0-1
 [3] dse1_2008.10-1                  tframe_2009.2-1
 [5] setRNG_2008.3-1                 lmtest_0.9-24
 [7] vars_1.4-5                      urca_1.2-2
 [9] strucchange_1.3-7               sandwich_2.2-1
[11] zoo_1.5-6                       MASS_7.2-47
[13] abind_1.1-0                     RcmdrPlugin.TeachingDemos_1.0-3
[15] TeachingDemos_2.3               RcmdrPlugin.HH_1.1-24
[17] HH_2.1-29                       leaps_2.9
[19] multcomp_1.0-8                  mvtnorm_0.9-7
[21] lattice_0.17-25                 RcmdrPlugin.IaeTlse_0.1-3
[23] RcmdrPlugin.Export_0.3-0        Design_2.2-0
[25] survival_2.35-4                 Hmisc_3.6-0
[27] xtable_1.5-5                    Rcmdr_1.4-10
[29] car_1.2-14                      JGR_1.6-4
[31] iplots_1.1-3                    JavaGD_0.5-2
[33] rJava_0.6-2                     fortunes_1.3-6
[35] ctv_0.5-2                       PolynomF_0.93

loaded via a namespace (and not attached):
[1] cluster_1.12.0 nlme_3.1-92    tools_2.8.1