Monday, October 29, 2007

A Thought on the Use of Factor Analysis

Derek H. C. Chen and Kishore Gawande have written a paper titled "Underlying Dimensions of Knowledge Assessment: Factor Analysis of the Knowledge Assessment Methodology Data". It refers to the data provided by the World Bank with the Bank's very nice Knowledge Economy Index toolkit.

The authors make the basic point, which I accept, that the 83 variables chosen to indicate the readiness of nations to develop knowledge based economies are highly correlated. The vast majority of information contained in the measurements of these variables can be communicated by using a smaller number of factors. They make the further point that the principle factors from a factor analysis do not convey that information in as intuitive a manner as other dimensions might, and thus that other combined variables might be better suited to enable the user to build knowledge.

It occurs to me that the purpose of the Knowledge Assessment Methodology simply descriptive, but evaluative. It seeks to enable its users to understand how ready a country is to develop socially and economically in certain directions. Comparing values of the Knowledge Readiness Index from one time to another is to help analysts to determine whether the country is moving appropriately in those directions.

The methods used by the authors discard information. How do they know that the information that they discard is not useful in the predictions that the database in intended to help users to make, and not simply noise. Unfortunately there is no real indicator of success in developing a knowledge economy that could be used for regression analysis.

Is it not the case that simply looking for combined variables that contain most of the information in a data set might result in a reduced set of information that looses key information for predictive purposes.

For example, might it not be that those countries which value scientific and technological information a little more than might be expected and that are a little more open than might be expected from their general level of development are exactly those nations that will move more rapidly towards a knowledge economy?

No comments: