unit root interpretation Please

[rselva123]

The result of Dickey fuller test shows Probability of 0.0044 (less than 0.05) but the t-statistics is -3.699 which is less than critical values. How should I interpret the result. Should I say 'The series has unit root? or should I reject the null hypothesis?
Thanks in advance

[startz]

You might want to post your output.

[rselva123]

Here is the output of study . Probability of 0.004 (less than 0.05) but the t-statistics is -3.699 which is less than critical values. How should I interpret the result. Should I say 'The series has unit root? or should I reject the null hypothesis? Thanks

Null Hypothesis: INDIAVIX has a unit root
Exogenous: Constant
Lag Length: 0 (Automatic based on SIC, MAXLAG=17)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -3.69932 0.0044
Test critical values: 1% level -3.44325
5% level -2.86712
10% level -2.56981

*MacKinnon (1996) one-sided p-values.


Augmented Dickey-Fuller Test Equation
Dependent Variable: D(INDIAVIX)
Method: Least Squares
Date: 03/03/12 Time: 12:04
Sample (adjusted): 2 499
Included observations: 498 after adjustments

Coefficient Std. Error t-Statistic Prob.

INDIAVIX(-1) -0.05469 0.014783 -3.69932 0.0002
C 1.254001 0.342806 3.658052 0.0003

R-squared 0.02685 Mean dependent var 0.006787
Adjusted R-squared 0.024888 S.D. dependent var 1.401791
S.E. of regression 1.384237 Akaike info criterion 3.492184
Sum squared resid 950.3919 Schwarz criterion 3.509094
Log likelihood -867.554 Hannan-Quinn criter. 3.49882
F-statistic 13.68495 Durbin-Watson stat 2.020148
Prob(F-statistic) 0.00024

[startz]

It's the absolute value of the t- that matters. You should reject.

[rselva123]

Thanks. Eviews gives confusing results. It's difficult to interpret.Having come upto version 7, it could refine its algorithm to give clear, consistent results.

[EViews Glenn]

I'd love to hear suggestions on how to make this clearer. We could spruce this up for EViews 8.

We currently provide the statistic and the p-value, as well as the critical values at the 10%, 5% and 1%. The latter are monotonically decreasing values with larger negative numbers associated with tighter confidence levels. So rather than thinking in terms of absolute values as in the usual t-stat, you should be thinking in terms of larger negative values providing greater evidence against the null.

In this case, the p-value is about .004 so that we reject. Re: your first post, note that we clearly state that the null is that there is a unit root so that rejecting means that you reject the null of a unit root.

[Davidkp]

Also, it's normal that the t value is less that the critical values as it is a one tail test. A one tail test can be positive or negative, you would reject a null hypothesis if (negative t-stat) < (negative crit value) or if (positive t-stat) > (positive crit value).

Correct me if I'm wrong?

[EViews Glenn]

In general yes, but in this application it's always the negative tail.

[hdhung]

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -2.885930 0.0572
Test critical values: 1% level -3.632900
5% level -2.948404
10% level -2.612874


In such case, should i reject the null hypothesis? Thanks.

[ekta1]

Hello Sir/Madam,

Please should absolute figures of the t-values be considered all time even if the t-value is positive and the critical values are negative?

Example my test results shows; t-value = 0.2768, critical values are; 1% = -2.8622, 5% = -1.979, 10% = -1.6337. p-value = 0.7882.

Please should I accept the null of a unit root.

Thanks, Kofi.

[gobbble]

You can definetely NOT reject the H0.