SAS Support Communities

Hi, An ARDL modeling process requires the variables to be of the mixture of I(0) and I(1) i.e. stationary and trend stationary respectively. My concern where it is not clear to me is that "What guides us" on model specification under Fixed Regressors to Either opt for 1: None 2: Restricted constant 3:Unrestricted Constant 4: Restricted trend OR 5: Unrestricted constant and trend   I have annual financial data, my D.V is FDI Inflows and my I.V are exchange rate, interest rate, inflation rate and GDP Growth rate In this case also which variables can I log transform to have same level of unit for easy interpretation? Is it recommended?   MY CONCERN WAS GENERATED AFTER A KEENLY READING THE FOLLOWING AS SUGESTED BY  @sbxkoenk    ARDL Models - Part I https://davegiles.blogspot.com/2013/03/ardl-models-part-i.html ARDL Models - Part II - Bounds Tests https://davegiles.blogspot.com/2013/06/ardl-models-part-ii-bounds-tests.html ARDL Modelling in EViews 9 https://davegiles.blogspot.com/2015/01/ardl-modelling-in-eviews-9.html See also here : "SAS Forecasting and Econometrics" board Autoregressive Distributed Lag Model https://communities.sas.com/t5/SAS-Forecasting-and-Econometrics/Autoregressive-Distributed-Lag-Model...   Many thanks, Appreciations!   Andy_20 ... View more

Hello there goods friends,  I am modeling factors for FDI Decision by analyzing macroeconomic indicators namely Inflation rate(%), GDP growth rate(%), exchange rate(TZS/USD), interest rate(%) as Independent Variables against FDI INFLOWS (Million USD), the dependent variable. The EXCHANGE RATE and Real GDP Growth rate series seem to follow a normal distribution but rest do not. The problem comes in transforming the data already in percentage into logarithm whose coefficients are interpreted as percentage change, can do that? What are possible consequences?  NB: All of the observations are positive but with some Variable having outliers.     I opted to use ARDL MODEL as the data are integrated of mixed orders I(0) and I(1)...! I now ask you help me if I can transform the data to esnure normality as well as reduce the Heteroskedasticity on the error term Thank you in advance, Blessings   ... View more