Time series regression/ARDL
Posted: Mon Apr 21, 2014 4:08 am
I will try and keep this as short as possible...I am a bit of a novice with stats as well, so bear with me!
I'm looking at the effects of government expenditure on economic growth, total factor productivity specifically, but only during a recession, so I have a very small sample size.I am running a regression with one dependant variable (TFP) and one independent variable (expenditure). As the title says I am using time series data.Originally I just did a linear regression, but then was worried about my variables being nonstationary as well as having no cointegration,causing spurious correlation. I then ended up having to take first differences to satisfy the augmented Dicky-Fuller statistic. My textbook then stated that I should take the first differences and use the Autoregressive Distributed Lag Model for my estimation. I then spent hours scouring the internet and Eviews, trying to find out how to compute it as well as trying to decide how many lags to use. I then began reading my textbook page by page (which was torture!), when I realised all of the examples in the book have at least 2 independent variables...
My questions are: do I still have to use the ARDL model with only one independent variable? If yes, does anyone know to do this in Eviews, and even if you don't, how do I find out how many lags to use ?
Tl;dr Can you use the ARDL model with only one x variable and a very small sample size?
I'm looking at the effects of government expenditure on economic growth, total factor productivity specifically, but only during a recession, so I have a very small sample size.I am running a regression with one dependant variable (TFP) and one independent variable (expenditure). As the title says I am using time series data.Originally I just did a linear regression, but then was worried about my variables being nonstationary as well as having no cointegration,causing spurious correlation. I then ended up having to take first differences to satisfy the augmented Dicky-Fuller statistic. My textbook then stated that I should take the first differences and use the Autoregressive Distributed Lag Model for my estimation. I then spent hours scouring the internet and Eviews, trying to find out how to compute it as well as trying to decide how many lags to use. I then began reading my textbook page by page (which was torture!), when I realised all of the examples in the book have at least 2 independent variables...
My questions are: do I still have to use the ARDL model with only one independent variable? If yes, does anyone know to do this in Eviews, and even if you don't, how do I find out how many lags to use ?
Tl;dr Can you use the ARDL model with only one x variable and a very small sample size?