Forecasting is a term used in economics to mean the estimation of the future state or position of a variable or a set of variables. Forecasting normally comes in handy in times of planning and making decisions for the organization about the variables in question. For instance, it could be used in determining the future rates of inflation of an economy, interest rates of investments, and the demand for products among others. As such, the methods used in forecasting should be very accurate and reliable since the outcome is highly depended on to make sensitive decisions of the organization.
Among the many methods of forecasting, regression analysis is commonly used. The regression could either be linear or multiple (Arsham 1). As for linear regression, the analysts normally take the variable to be forecasted for instance sales, and develop a relationship between this variable and another one independent variable. This relationship is then used to make a forecast. Therefore, when collecting the data to be used in linear regression, a sample of the variables, which forms the subset of the population, is normally used. This is because a regression analysis of the whole population is unrealistic because of the inability to get full information of each.
Therefore, the problem of representation comes about because of sampling since the selected sample could not be representing the whole population thus giving the wrong inferences. Once the sample is picked, Matrix Algebra techniques could be applied to solve the linear equations created in the relationships (Arsham 1). One shortcoming of this method is that it requires more than two equations, as well as, two unknowns for the equation to be computed. This means that data collected and analyzed using linear regression can only be used to compare two variables to make a forecast. The limitation to only two variables is a potential problem as the inferences are drawn are not comprehensive. However, several computer packages could be used to solve the linear equations.
Another shortcoming of linear regression is that the predictors of the variables are linearly independent such that they cannot be used on other variables. This, therefore, means that for every forecast using different variables, you have to create different relationships for the linear regression (Arsham 1). This is very tiring especially because other methods could be used to overcome this shortcoming. Finally yet importantly, just like in most forecasting methods, linear regression also has to deal with the problem of error margin, hence making the degree of accuracy to decrease.
Works Cited
Arsham Hossein. Time-Critical Decision Making for Business Administration. 1994. Web.