Recall issues that surface from omitted variables. They cause biased results, including biased coefficients and error term. That is where your missing variables are accounted for in regression analysis.

The very first thing to do, and most important, is to review economic literature to determine what possible variables are missing. Any missing variable will cause biased results within the coefficient(s) and error term depending on the degree of correlation(s) among the variables in question.

Thanks for your inputs. But suppose that the dependent variable is a general one like sales which is affected by a huge number of internal and external factors. Now if I want to study the extent of effect of say corporate governance or liquidity policy then how can just find that.?

The CLRM gives a solution to this problem - the assumed error term u in the functional forms of the bivariate or multivariate regression equations actually accounts for the omitted variables or the ones the researcher is not interested in the study. The stochostic Y and the independent X ( or X1, X2 etc) and the random error term provide the researcher the parameters for estimation through the OLS or any other model. Here the researcher chooses the X1, X2 because he is interested in their impact on Y; but he omits to consider the other variables ( may be necessary or only a nuissance) but their effects are subsumed in the error term as per the theory of the CLRM and the properties of the stochostic error term included in the functional form. The more the variables, lesser the degrees of freedom, their correlation among them affecting the precise estimate, not being able to pin pointedly find out the effect of the particular independent variable on the dependant Y and finally making the model more complex and less effective in getting the result. Hope this clarifies,

Dr Durairajan

As previously discussed: firstly you should study papers that are related with your area of study, secondly you should read a book about econometrics i.e. How to create a model, what is the nature of data, correlation matrix and so on.

Suppose we are hypothesizing that a set of macro economic and firm specific variables are affecting a specific financial variable of firms. And this question some how made a lot of sense to all those who have answered. My head is spinning at the use of "firms" n two different places in this sentence where we are supposed to hypothesizing and trying to get clarification, etc. Firm specific variables and variable of firms - I don't even understand this so how can I answer the question! Jargon, jargon, jargon! Just kidding, folks.

I would like to call attention to what I have called a generalization of the concept of the work function, first conceived by Einstein in 1905 to explain photoelectricity. If E = hf is the energy of the photon, the MAXIMUM (kinetic) energy of the electron that is produced by the photon will be given by K = E - W = hf - W where "W" is the work function. This is the energy that must be given up to bring the electron out of the metal and overcome the complex forces that bind it to the metal. Instead of trying to calculate W, Einstein simply wrote K = hf - W where f is the frequency of light and h is the Planck constant. Hence, the K-f graph will be seen to be a family of parallel lines, if we perform experiments with different metals, each with its own value of W.

In the financial world, I have shown, after studying the data for literally hundreds of companies, that an exactly similar law describes the profits and revenues relation which can be written as y = hx + c where x is revenues and y is profits and the nonzero intercept c is akin to the work function conceived by Einstein. If we prepare a graph of revenues versus profits for a company (e.g., Apple, Google, Microsoft, to name a few) we will see a nearly perfect straight line with a fixed slope. Also, we see the data moving along parallel lines, as I have shown for Microsoft, and also for the new post-bankruptcy General Motors.

Thus, the answer to the question posed is the "work function" takes into account the variables that have not been fully accounted for. The articles alluded to here can be found at www.scribd.com where I have uploaded two files with a detailed bibliography (Part I and II) giving links to the articles and the titles to each in full. Here's Part II link, which gives the link to Part I.

http://www.scribd.com/doc/147955814/Bibliography-II-of-V-Laxmanan-Articles-on-the-Extension-of-Planck%E2%80%99s-and-Einstein%E2%80%99s-Ideas-Beyond-Physics-with-Examples-from-the-Observations-on-Finan