1
$\begingroup$

Let's say I have a pure random walk:

library(fUnitRoots)
library(urca)

set.seed(1130)
rndwlk1 <- filter(rnorm(1000,0,1),c(1),method="recursive",init=c(1))
plot(rndwlk1)

and I want perform a unitroot test with trend using ur.df, I would define my series as such:

$y_t = c + \delta t +\rho y_{t-1} + \epsilon_t $

summary(ur.df(rndwlk1,type=c("trend"),lags=0))

## Call:
## lm(formula = z.diff ~ z.lag.1 + 1 + tt)

## Residuals:
##     Min      1Q  Median      3Q     Max 
## -3.7054 -0.6353  0.0402  0.6363  2.9240 

## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)   
## (Intercept) -0.186081   0.077171  -2.411  0.01608 * 
## z.lag.1     -0.017740   0.005983  -2.965  0.00310 **
## tt          -0.000619   0.000233  -2.657  0.00802 **
## ---
## Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

## Residual standard error: 0.9726 on 996 degrees of freedom
## Multiple R-squared:  0.00875,    Adjusted R-squared:  0.00676 
## F-statistic: 4.396 on 2 and 996 DF,  p-value: 0.01256


## Value of test-statistic is: -2.9648 3.8345 4.3962 

## Critical values for test statistics: 
##       1pct  5pct 10pct
## tau3 -3.96 -3.41 -3.12
## phi2  6.09  4.68  4.03
## phi3  8.27  6.25  5.34

Can someone tell me what the test-statistics mean? As far as I understand the first one is the ADF-statistic for $\rho$ corresponds the -2.648 is the t-statiscics for $\rho$. What are the two other statistics?

The model with trend is not the right model, but I would like to test it against a model with drift. I have figured out that the F-statistics gives the results for the test of a unit root with drift. So:

H0 = $\rho$ = 1 , $\delta t$ = 0. But the p-value seems to be from a regular F distribution. How can get critical values for the F statistics?

Update

The p-value of the F-statistic is indeed from the regular F-distribution. pf(4.396,2,996,lower.tail=FALSE)

#0.01256662

Does anyone know where can we find the Dickey fuller critical values for the F-statistics on R?

$\endgroup$
1
$\begingroup$

First of all, the function ur.df will estimate the following equation:

$\Delta y_t = c+ (\rho-1)y_{t-1} + \delta t + \epsilon_t $

The value of the test-statistics:

## Value of test-statistic is: -2.9648 3.8345 4.3962 

représents the critical values for the following $H_0$.

|----------------+----------------------+--------------------|
|        -2.9648 |               3.8345 |             4.3962 |
|----------------+----------------------+--------------------|
| $(\rho-1) = 0$ | $\rho-1= c=\delta=0$ | $\rho-1= \delta=0$ |
|----------------+----------------------+--------------------|
  • What are the two other statistics?

So the two other values are critical values for F-Test.

  • How can get critical values for the F statistics?

The corresponding critical values are given in the table below the test statistics. the tau3, phi1 to phi3, are the notations DF used in their paper. I found the table and the explanation in Applied Economic Time Series, Walter Enders. Here is what the notations represent. Where $\tilde{\rho}$ = $\rho-1$

|--------+---------------------------------+-------------------------------------|
| <6>    | $H_0$                           | Interpretation                      |
|--------+---------------------------------+-------------------------------------|
| tau3   | $\tilde{\rho}= 0$               |                                     |
| phi1   | $\tilde{\rho}= 0 et c = 0$      | random walk without drift           |
| phi2   | $\tilde{\rho} = c = \delta = 0$ | random walk without drift and trend |
| phi3   | $\tilde{\rho} = \delta = 0$     | random walk without trend           |
|--------+---------------------------------+-------------------------------------|
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.