It’s a very common misconception that economists just study the economy. That is only one aspect of economics. There is a huge amount of economic research which concerns itself with health. So for this week’s post I thought I’d write a piece which touched on this. Specifically, I am going to show how to run a cox regression.
This is a quantitative estimation technique which examines the effect which several variables have upon the time it takes for a certain event to occur. The event in this case is the death of a musician from drug consumption. We are going to see how taking a drug will effect the amount of time it takes for a musician to die.
The Stata data set I will be using can be found here ->Musicians and Cox Regression and is free for all to download. Though it should be noted that this is fictional data and doesn’t pertain to any real individuals. The do file I will be using can be found here -> Do file for musicians and cox regression. Simply download the data file and paste the do file into your Stata do file accordingly.
The results of the cox regression can be seen below. A note worthy point regarding the cox regression is that it produces hazard ratios. These indicate how much more quickly the event they are testing for will occur given an increase in the explanatory variable. The hazard ratio value of 0.258 indicates that a musicians death will occur 0.258 times faster with every unit increase of drugs they take.
Obviously it is hard to see the effect which increased drug consumption has on the time it takes for a musician’s death to occur just from numbers. This is why it is necessary to generate a Kaplan-Meier graph. This graph will show graph of decay in the amount of time it takes for a musician to die given the dosage of a drug that they take.
As can be seen above the musicians who take the highest dosage of drugs (10mg) die just before 25 units of time, the musicians who take the next highest dosage of drugs (5mg) die just after 30 units of time, and the musicians who take no drugs at all (0mg) last until nearly 40 units of time. This type of quantitative testing is usually used by health economists who are doing feasibility testing for incoming drugs which are new to the pharmaceutical market.
By Daragh O’Leary