I have mentioned the gld gdf incorrect axis in the comment of ur response.
But i also had a doubt on the critical value selection i.e. selecting the probability of 90 or 95 or 99 %
while adf critical value number is fixed…
and if the test statistic falls between
90 and 95 : do we take critical as 95 and accept null or critical as 90 and reject null
95 and 99 then too, how do we decide whether to reject or not?
Also if test stat falls exactly on
critical of 90 or 95 or 99 how to proceed?
The more negative the test statistics value the better it is.
Considering the below table, if the t-stat value is -2.95, then with 95% confidence you can say that the time series is stationary. If the t-stat value is the same as the critical value and if you are a consverative trader then you should consider the time series not to be stationary. As most likely, if new data points come in the t-stat value can become greater than the critical value.
| Probability | Critical Value |
| 90% | -2.57 |
| 95% | -2.87 |
| 99% | -3.44 |