ADF- trend and intercept

[fatyma]

Hello, i wanted to ask that when we apply ADF test on eviews, it gives us 3 options, none, intercept and trend, intercept. How can we know that wether to go for trend and intercept or none. Is there any command on eviews through which we can know about our data generating process?
Thanks, looking forward to replies

[econometricsI]

well you would start off with a trend and intercept model. But the problem I have is how to decide if the trend or intercept are significant or not.

[sama]

True, you should start testing for trend and intercept at first, BUT you should do this at level (i.e. no differences taken). in this case the Hypotheses are:
H0: the trend coefficient = 0 => the data has no trend and needs to be differenced to make it stationary
H1: the trend coefficient < 0 => the data follows a Trend Stationary Process (TSP) and you need to include the "time" variable in the regression model instead of differencing the data.

So if your p-value is less than 0.05, then you have enough evidence to reject H0 and to consider that you data is TSP.
Else, you need to test for the 1st difference, without the trend.
the Hypotheses are:
H0: the trend coefficient = 0 => the data needs to be differenced to make it stationary
H1: the trend coefficient < 0 => the data is stationary

P.S: Always try to confirm your test results with the graphs and the correlograms.

Regards,

[sama]

correction:

for intercept+trend:
H0: the coefficient of Y(t-1) = 0 => the data needs to be differenced to make it stationary (regardless of the trend)
H1: the coefficient of Y(t-1) < 0 => the data follows a Trend Stationary Process (TSP) and you need to include the "time" variable in the regression model instead of differencing the data.


for intercept only:
H0: the coefficient of Y(t-1) = 0 => the data needs to be differenced to make it stationary
H1: the coefficient of Y(t-1) < 0 => the data is stationary

[tree_6]

So if your p-value is less than 0.05, then you have enough evidence to reject H0 and to consider that you data is TSP.
Else, you need to test for the 1st difference, without the trend.
the Hypotheses are:
H0: the trend coefficient = 0 => the data needs to be differenced to make it stationary
H1: the trend coefficient < 0 => the data is stationary




I got cnfusion about why you test for the 1st difference, without the trend?