# What are the statistical tests I need to perform before I regress GDP on multiple variables?

I'm using data over 10 countries for the period 1975 to 2010. My objective is to check if the chosen regressors have significant effect on GDP. I have very basic knowledge of econometrics which I learnt in my first year of college. I was introduced to simple linear regression model with k variable which involved estimation of OLS, calculations of standard error, hypothesis testing for the slope coefficients. All of these were done with the assumptions of no perfect multicollinearity and the random disturbance term U follows normal distribution with constant variance. I have read a few papers on similar topic and they perform tests like unit root and Granger causality but I don't understand them. So what are the topics I need to learn to be able to perform a test of significance for a panel data? Also, what would be an easy to understand book with stata examples for the suggested topics?

• Theres a lot to truly understand how to model with panel data. If you have the time and are interested enough, Cameron & Trivedi’s “Microeconometrics Using Stata” is a fantastic book that starts from the basics and works through just about everything you would need to know. All examples use Stata input and output and are completely reproducible as they use common datasets you can load right into your program with a single command. Highly recommend – Brennan Sep 4 '19 at 20:19
• Strictly speaking, it is not a panel -data. It is a time series cross-section data. unit root and Granger causality are not meant for significance testing. You have multiple unstructured objectives that need refinements. – Subhash C. Davar Feb 17 at 0:23