How to Estimate and Interpret a Linear regression in Eviews
Learn how to estimate a linear regression in eviews
Welcome to a new Eviews tutorial. In this post, you will Learn how to estimate a linear regression in eviews. The tutorial is simple and basic, but it is a good starting point for those new on the subject. I hope you find it useful!
By watching this tutorial, you will learn:
- The steps to estimate a linear regression in EViews
- How to read and interpret the estimation output in EViews
what is a linear regression?
Linear regression is a linear approach to modelling the relationship between one or more explanatory variables. When there is one explanatory variable, we are dealing with a simple linear regression. When there are diverse explanatory variables, we are dealing with a multiple linear regression.
things to consider when estimating a linear regression
Although estimating a linear regression in EViews can be simple, understanding the output and determining whether we are in the presence of a spurious regression is not that easy. We want to ensure that the model estimated is appropriate and can help us to understand the effect of the independent variable on the dependant variable.
understanding the linear regression estimation output
coefficient column (1)
The coefficient column displays the estimated coefficients in the model. The estimates are computed by OLS
In our model, since we are using logs and displaying the results as elasticities, we can conclude that: If the GDP of Brazil increases by 1%, the GDP of Argentina will increase by 0.49% (half percent).
standard errors column (2)
The standard errors show the estimated errors of the coefficient estimates of the model. If the deviations are large, the accuracy of your coefficients estimates is very low. Your model will not be reliable.
T-Statistics column (3)
The t-statistics will allow us to conduct diverse hypotheses, such as the individual significance test. The null hypothesis of this test indicates that the value of the coefficients estimates is 0. But, what does that mean? If we confirm the null hypothesis, we are saying that the variables do not help to explain the dependant variable (in other words, our explanatory variable explains 0).
TIP: When the t-statistics are very large (i.e., T-Statistics = 17) our coefficient estimates will be highly significant. Large T-statistics values are commonly associated with spurious regressions.
probability column (4)
The last column (probability) stands for the p-value. The p-value will allow us to reject (or not) any given null hypothesis.
In this case, the hypothesis tested is the individual significance test.
If p > 0.05, at the 5% significance level, our explanatory variable (often called independent or RH variable) is not significant. Including this variable in the model will not be appropriate as it will not help us explain the dependant variable.
If p < 0.05, at the 5% significance level, our explanatory variable significant. Including this variable in the model will be appropriate as it will help us explain the dependant variable.
In our Model, we can see that the constant term and the log of Brazil GDP, does help to explain Argentina’s GDP (since p<0.05).
R - Squared (5)
The R-Squared takes values between 0<x<1, and reflects how well the explanatory variables predict the dependant variable. In other words, it can be explained as the fraction of the variance of the dependent variable predicted by the explanatory variables. If the R-squared is closer to 1, your model has a good fit, while if the model is closer to 0, the model does poorly.
Adjusted r - Squared (6)
The adjusted r-squared provides a similar information than the r-squared, however, it is adjusted by the amount of explanatory variables. If you estimate a model with many explanatory variables, the adjusted r-squared will be lower in comparison to the r-squared. We use the adjusted r-squared for models where you incorporate many explanatory variables.
s.e of regression (7)
The standard error of the regression is computed as the square of the variance of the model.
sum squared resid (8)
The sum squared residuals are calculated and provided as an additional information which can be useful for diverse hypothesis testing.
Download the dataset for free and replicate the content covered in the video
Download the Data Set for Free (direct download, no adds).
Follow me on social media and stay tuned to tips, free files, promotions and new videos !
Thanks for your support!
Did you know you can buy my EViews workfiles, stata DO files and OVerleaf templates?
Check out my store to find out the available files that will help you in your reserach