SAS software

In USA, I see SAS is a popular software for data analysis, even EVIEWS is not popular here in which I have command.

I also know SAS but not as much as I have on EVIEWS.

Anova

Anova is appointed when you want to find out mean difference between variables. Example, whether there is a mean difference in weather among five states or whether there exists a mean difference in age among five countries etc.

Anova is mostly used in SPSS

Forecasting

When I go for forecasting any variable, I normally use few methods. ARIMA (Auto-regressive integrated moving average method) is one of them where four steps are involved. They are identification meaning how many would be AR and MA. Afterwards I estimate the model using appropriate AR and MA. Then I go for diagnostic checking of the model meaning that whether the residual of the model is stationary or not. If it is stationary, meaning that our ARIMA model is perfect. Finally I go for forescting of the variable.

But in ARIMA make it sure, variable must be stationary. Some time ARIMA model is called Box Jenkins Methodology. It is one of the complex models in econometrics analysis.

The complicated part of ARIMA model is to find out approproaite number of AR and MA. There is some guideline such as shape of the ACF and PACF distribution from there we can know how many would be AR and MA>

Second appropach I use for forecsting is VAR model. It is easy to use. Specially under EVIESW it is very simple. But make sure data must ne statioanry.

In time series operation, data must be stationary.

Time series and cross section data

Time series data means data has been collected over time such as from 1980 to 2009 while cross section data means data has been collected at a particular point of time. Such as, 400 managers perception in a particular time such as year 1980.

When we run the regression line, the residual of the model may be serially correlated (which is not desirable) if the data or variable is time series type. Normally heteroscadasticity does not arise when the data is time series type.

But when we run the regression line, the residual of the model may be heteroscedastic (which is not desirable)if the data is cross section type. Normally serial correlation or auto-correlation does not appear when the data is cross section type.

Solution for Heteroscedasticity

Suppose you are running the regression line where Y is dependent variable while X1 and X2 are independent. After runing the regression model, you see that residual of the model is heteroscedastic, which is not desirable.

To solve the problem, there are many ways. One of the ways, convert all variables into natural log, that is lnY, lnX1 and lnX2. Then run the regression line agin. You may find that residual is no longer heteroscedastic. It has become homoscedastic whioh is desirable.

Which software is the best to solve time series model?

EVIEWS is the best software to solve time series model, especially problem dealing with stationarity of data or unit root testing, regression model, residual management, restricted and unrestricted VAR, structural VAR, ARCH, GARCH, E-GARCH model.

It is user friendly and easy to execute. You just to go HELP of this software and you will find all the steps necessary to execute the software. HELP also explains statistical issue regarding the model.

What is the main task of an econometrician?

The main task of an econometrician is to manage the residual of a model, meaning that model must be free from heteroscedasticity, non-normality of residuals and serial correlation.

In other word, the regression model must be white noise

What are the best books in Econometrics?

Books written by Damodar Gujarati, Pindyck and Green.

For beginners, start with Basic Econometrics by Damodar Gujarati, McGraw Hill.

For advanced level, Follow Pindyck, Green, Johnston etc

Should we reject null hypothesis at 10 percent?

Yes, we can reject null hypothesis at 10 percent level but the probability of committing type I error would be 10 percent. So we should not reject.

When a variable is significant?

A variable, suppose X is significant at five percent level when the p-value associated with t-statistics is less than 5 percent.

Suppose, p-value is 0.049 meaning that X variable is significant at 4.9 percent level.

In other way, we can say that, if we reject the null hypothesis, the probability of committing type I error would be 4.9 percent. So we can reject null confidently as the probability of committing type I error is very small.

What is econometrics?

Econometrics is an amalgamation of statistics, mathematics and economic theory.

Normally an economist runs the econometrics model to prove or disprove a particular phenomena or theory.