Logistic regression is used to convert the qualitative information into the quantitative one. Search engine algorithm has lots of qualitative aspects. Example:
1. Presence of physical existence of a company (1 if yes, 0 if no) – Engine do provide values on the basis of the results found.
2. In coming links relevancy (1 for yes, 0 for no)
and so on….
The values in the brackets are quantitative data found against the qualitative aspects for which we can fit a suitable regression model to find out which algorithms are more important (the co-efficients, t values, R etc.) to raise a website in the SERPs.
In the next step we can move further for the web usability (age, sex, sectors, regions, languages etc) so that predicting the algorithms, usability will provide more spaces to marketize the products, building brand online, diversify the products (both vertically and horizontally).
Q. Could we build another web analytics parallel to the Google one based on Regression and Statistical analysis?
A. Yes, we can do that. In that the programming tools will be the complementary support for analysis. And I suppose SAS would be of great help.