, , , , , , , , , ,

Yesterday I had two conversations about regression modelling. The first was with an esteemed fellow enthusiast and the other was with someone who knew nothing about regression modelling beforehand. I am a self-confessed regression modelling enthusiast. The various models have different applications depending on what relationship the researcher wishes to investigate.

My fellow enthusiast is a fan of the logistic regression, but did not know about multinomial logistic regression, much to his shame, but I feel that the logistic regression is limited in its application – usually confined to health economics.

I love the two-step least squares model, or 2SLS, as it is more in line with my professional work and personal interests. 2SLS is used, primarily, for the investigation of a relationship between something (x) and price (£).

An example I used 2SLS in yesterday was why iPad and iPhones command such a high price. The reason being is that there is a relationship between demand (x) and the price (£). High demand of the Apple product allows Apple to use 2SLS to set the price at the maximum tolerable level before demand would start slipping. Similarly, if the Apple product had low demand it would have to set a low price in order to increase demand. Of course, there are more statistical tools in play when setting the price of a product, but 2SLS is, I believe, instrumental in that process.

This morning I was watching BBC Breakfast and one of the articles was about increasing the price of fizzy drinks to slow down the obesity rate as, using another regression model, probably a logistic model, researchers had linked fizzy drinks to be a cause of obesity and related, and unrelated, health problems. By using 2SLS researchers are able to suggest an increase in price to a level which would show a drop off in demand. Manufacturers of such products are, naturally, not going to be happy with such a suggestion but the effects should start being noticed.

I have also begun reading an Adam Smith Institute report on sin taxes, for which I will soon critique, and from what little I have read I can only assume that some sort of regression modelling was used in the creation of the report. Just like regression modelling will feature in my critique.

I raise the issue of regression models here as they are often an overlooked tool which affects our everyday lives. Pay attention to the models, see the models and love them.

For further reading on regression models, I direct you to this manual from IBM: IBM SPSS Regression 20