hi dear Udari ,

in ARDL you should include intercept in your model,

if it was nonsig you should exclude intercept.

I hope this could help

It depends on the nature of your variable. I suggest you read some econometric books to find out your answer.

Select the variable column in excel and see if there is any trend. Than apply the ADF test accordingly.

Thanks all for your valuable contribution.

You should include constant and trend. Do not worry about looking at the graph because we generally misread them.

you have to do and see. first visually check (line graph) check whether trend pattern available or not , check intercept also

then check significance of the each term

all depends on the data series.

First of all, you have to check your data through plotting, if which has trend and intercept you could use the trend and intercept in ADF test. Therefore, your ADF test depends on the natures of  your data  

@Sivarajasingham and Ahamed

Assume 1 and t are included and the t in statistically insignificant, what is the null hypothesis associated with the t? I have no problem with plotting variables although my preference is starting with a full model.

Hi Udari,

Testing for stationarity can be quite tricky. It's best to start off running the test with trend and intercept and then evaluate the significance of the components. Plotting can indeed help with the initial assumptions but I wouldn't trust it too much (you cannot observe a zero trend or intercept, for instance). In addition you could run tests with only intercept and trend and without both, respectively, and then compare information criteria (AIC, BIC, LL). In the best case, you won't have to worry about much as the components are either significant or the information criteria point towards the same model. In any cases of doubt you might want to err on the side of non-stationarity. In this respect, it's sometimes useful to run a KPSS test in addition to the standard ADF test, as it tests the opposite null, I.e. stationarity. Hope this helps. 

it depend to your variables. intercept is better to be in the model.but the relation to intercept,  if variables have been trend, should include trend in your model.