Q1. What the lag length criteria? How can I select legs for a model?

Ans: Lags in a regression (panel data, time series) are selected based on AIC, BIC/SIC, and HQIC. The conventional rule is to select a model with specified lags with lowest values for AIC, BIC/SIC, and HQIC.

Q2. How can I run ARDL in Eviews?

Ans: You can once-left-click on the dependent variable (DV), press Control (CTRL) key on the keyboard and click on other independent variables (IVs), then right click on selected variables, Click on Open, Click On As Equation. It gives a small Eviews window. Select the

 Type/Model/Method as ARDL. 

It gives another Window. 

Select values for Maximum No of Lags for DV and IVs.

 The maximum lag length is generally 4 for macroeconomic variables (you can see this from literature as I have seen it in the literature on the environmental Kuznets curve and emissions the for a paper). 

Then click on OK to estimate the model and produce the results.

Q3. How to test a variable is significantly affecting another variable?

Ans: Significance is a relative term. There is a convention and in most economic and financial research, the confidence level of 99% and other related fields the confidence level of 95% uses a benchmark for testing significance of the effect of one variable on the other. The effects are tested through regression against correlation which is used to test the only existence of relationships. A variable is said to significantly affect another variant if the test statistic which is calculated using the given formula in books is greater than tabulated value of the given statistic for the given confidence level. Alternatively, you can see the p-value which is calculated as the probability for (calculated statistic>tabulated statistic). If the p-value is less than 1-confidence level, then reject the null hypothesis. Where the null hypothesis is the one variable does not affect other variables.

Q4. What is the difference between correlation and regression?

Ans: Regression is the term used for the testing the Effect of one variable on another variable and correlation is the term to test the existence of relationship between the two variables and two distinct terms in statistical inference because effects is a directional link between two variable where we can specify if one variable affects the other variable (the direction of effect is from one variable to another variable) while the existence does not take this direction into effect. It can be one way or the other (this not uncommon to see one or the other way of the effects or both ways) but we will only see if a change in one variable is related to change in another variable through common variations/deviations. The effects are tested through hypothesis on Beta and hypothesis on correlations is tested through R or r (notation for correlation).

Q5. How to run Unit Roots using Eviews?

Ans: This question is answered for both time series and panel data equally as the steps are same to estimate models because Eviews stored your data differently but clicking and producing the results for Unit Root are same.

First, select the variable that you wish to test for Unit Root. Then click on Open, Then click on As Graph, a small window opens. See at the left upper corner. There you will see View menu. Click on it and Click on the Unit Root test. Eviews will give different options for both TS and PD. Select the relevant check boxes and make the tests. The Null Hypothesis is given in the first line mostly or make sure you note Null from literature. See the decision-making rule is simple. If the p-value is less than conventional 5% or 0.05, reject the null hypothesis.