Step-By-Step Guide on Interpreting your Eviews Regression Output

When you perform a basic multiple regression analysis on Eviews, your output will generally be of the form of the frame below:

Output above is divided into three parts:
1) General Information
2) Relative Statistics
3) Global Statistics





The first part contains general information about the nature of the data, method(s) of analysis and date and time the analysis was performed.

A closer observation will reveal the following the in the first part:
1) The first line informs us that the dependent variable is GFCF (Gross Fixed Capital Formation).
2) The second line identifies the method of analysis as ordinary Least Squares.
3) The third line tells us the time and date the analysis was performed.
4) Fourth line informs us that the sample was collected for the period covering 1981 to 2012 and
5) The fifth and final line in part one says the number of observations (sample size) is 32 (years).

Part two is termed relative statistics because the output in this part of the result relates to the specific variables used in the study.

This part is made up of five columns namely:
1) Variable
2) Coefficients
3) Standard Error
4) T-statistic and
5) Prob

The first Variable identified is tagged C which means constant. this is the value dependent variable will take in the absence of interaction with the next two variables identified in th study. 

Anyway, don't worry your head abou this first necause it makes little sense on interpretation so the general opinion among researchers it to ignore it and concentrate on the rest of the variables. That being case, we go to line two with the variable name being DOMDEBT (Domestic Debt). 

From the output, the coefficient of DOMDEBT is 0.991086. this value means that there is a positive relationship between DOMDEBT and GFCF with the implication that every unit increase in DOMDEBT is predicted to be accompanied by 0.991086 units increase in GFCF. 

The standard error of the regression coefficient is 0.046656. This value identifies limit of error is expected to be inherent in the result. From the result, the expected error in the result is plus or minus 4.67%

The next is the t-statistic which gives a reading of 21.242. Here, you are expected to compare this value to the critical t-statistic table in the back of every standard statistics textbook. If the output t-statstic (the one above) is greater than the critical t-statistic in your textbook, then the null hypothesis is rejected if not, the null hypothesis should be accepted.

Considering that our output t-statistic above is greater than the critical t-stat of ---- we reject the null hypothesis  and conclude that DOMDEBT significantly affects GFCF.

The Prob of the above t-statistic is next which indicate that indeed DoMDEBT is statistically significant @ 1% confidence level. 

Now follow the same process to interpret the next line (EXTDEBT), what did you arrive at? 

Use the comment box at the end to share your answer with us.

part three is termed global statistics because the result in this part do not deal with specific variables but looks at the goodness-of-fit of the model specified to test the data.

In part this of the output, the interpretation will centre around the starred items as they are the relevent ones for a basic multiple regreession anlaysis.

Here we first look at R-Square (coefficient of determination) and this gives the value of 0.939810. This value implies that DOMDEBT and EXTDEBT can be relied on to explain 93.98% of the variations in GFCF.

The second starred item Adjusted R-Square is basically the same as R-Square above R-Square adjusted for error in data. The result also indicate that the level of error is negligible (approximately 0.42%).

The third item is the Durbin-Watson stat which is a daignostic statistic which tests for the presence of auto-serial corrleation in the data. 

As a rule of thumb, Durbin-Watson statistic runs from 0, 1, 2, 3 and 4. A value a close as possible to the central value of 2 indicates the absence of auto-serial correlation while any value far to the left or right indicate the presence of negative or positive auto-serial correlation with the implication that the results may not be entirely reliable.

Here, the result indicate the presence of negative auto serial correlation.

We'd appreciate your responses and observations, please use the comment box below of email optistrat.info@gmail.com

Go to Homepage to view a list of our well researched, fully supervised and approved Accounting and Banking and Finance Project Topics and Materials.