Saturday, February 03, 2007
On Conservatives, Liberals and.....Hypothesis Testing
What makes a person conservative or liberal? I don't know.
But the theory of Hypothesis Testing does help us gain insight into a person's political predilections. Gobsmacked, eh? Read on.
Suppose, for the sake of argument, that youth from the Middle East/N.Africa are more likely to resort to terrorism than any other race. To strengthen this supposition, we have to reject the hypothesis that terrorism finds no special favour among any specific group of people.
H0 : Youth from the Middle East/North Africa are no more inclined/indifferent towards Terrorism than youth from any other part of the world.
The alternative hypothesis is that youth in certain parts of the world are indeed more likely to turn into terrorists. Hence, racial profiling can help apprehend potential terrorists and debar entry to dubious immigrants.
For instance, assume that 5% of the youth wanting to immigrate to US are of Middle Eastern/N.African extraction. Going by the Null Hypothesis, in any sample of immigrant terrorists, the proportion hailing from M.E/N.A shouldn't exceed 5%. Right?
Now, consider a random sample of immigrant terrorists wherein 20% are Arabs/N.Africans and the probability of more than 20% of a random sample of immigrant terrorists being N.Africans/Arabs, given that the null hypothesis holds, is 5%.
Does this figure justify racial profiling? This is where your political predilection comes into play. A liberal might be wary of rejecting a true hypothesis (Type I error) and may be unwilling to disregard it even if the figure was as low as 1%. A conservative rightwinger on the other hand, would not want to run the risk of accepting a false hypothesis (Type II error). He may be tempted to ditch the hypothesis for any figure less than 15%!
This has to be one of the great debates in public policy. What is the most appropriate significance level? Should policy makers opt for a high significance level and aggressively pursue racial profiling to weed out the slightest possibility of terrorist infiltration at the expense of civil liberties? Or should they refrain from 'profiling' altogether, and thereby run the risk of terrorist attacks?
The answer would vary depending on whether you're a fan of George Bush or Michael Moore ;)

What makes a person conservative or liberal? I don't know.
But the theory of Hypothesis Testing does help us gain insight into a person's political predilections. Gobsmacked, eh? Read on.
Suppose, for the sake of argument, that youth from the Middle East/N.Africa are more likely to resort to terrorism than any other race. To strengthen this supposition, we have to reject the hypothesis that terrorism finds no special favour among any specific group of people.
H0 : Youth from the Middle East/North Africa are no more inclined/indifferent towards Terrorism than youth from any other part of the world.
The alternative hypothesis is that youth in certain parts of the world are indeed more likely to turn into terrorists. Hence, racial profiling can help apprehend potential terrorists and debar entry to dubious immigrants.
For instance, assume that 5% of the youth wanting to immigrate to US are of Middle Eastern/N.African extraction. Going by the Null Hypothesis, in any sample of immigrant terrorists, the proportion hailing from M.E/N.A shouldn't exceed 5%. Right?
Now, consider a random sample of immigrant terrorists wherein 20% are Arabs/N.Africans and the probability of more than 20% of a random sample of immigrant terrorists being N.Africans/Arabs, given that the null hypothesis holds, is 5%.
Does this figure justify racial profiling? This is where your political predilection comes into play. A liberal might be wary of rejecting a true hypothesis (Type I error) and may be unwilling to disregard it even if the figure was as low as 1%. A conservative rightwinger on the other hand, would not want to run the risk of accepting a false hypothesis (Type II error). He may be tempted to ditch the hypothesis for any figure less than 15%!
This has to be one of the great debates in public policy. What is the most appropriate significance level? Should policy makers opt for a high significance level and aggressively pursue racial profiling to weed out the slightest possibility of terrorist infiltration at the expense of civil liberties? Or should they refrain from 'profiling' altogether, and thereby run the risk of terrorist attacks?
The answer would vary depending on whether you're a fan of George Bush or Michael Moore ;)
Labels: economics, politics, statistics