Ideas diary

This page contains a running list of ideas. Some are just thoughts; others are things I'd like to explore more. If you're interested in talking about or collaborating on any of them, give me a holler.

(12/27/12) Encouraging the use of European style cars with poor acceleration: When I've rented small cars in Europe, I've gotten in the neighborhood of 40-50 mpg in non-hybrid cars. Why can't standard small American cars get this kind of mileage instead of the 30-35 mpg they often get? My theory is that it's the characteristics of the engines and that European cars are designed for good mileage and American cars for performance, because Americans won't accept the poor acceleration of such cars. Even among friends who are environmentally conscious, I've often heard complaints about cars that are slow to accelerate. So let's start a movement of Americans who want cars that have no pickup but get great gas mileage. The common argument is that cars that don't accelerate are dangerous when one needs to accelerate quickly, but the Europeans seem to do fine, and I've driven plenty of cars that don't accelerate well - you just have to plan for it and know that's how the car responds.

(12/27/12) Matching statisticians with applied problems: I've thought for a while that statistical talent and scientific need for statistical expertise are poorly matched. Too many statisticians are working on problems of little scientific/societal importance, while many important, and complex, analyses are done with little input from statisticians. So I'd like there to be a website for matching projects with statisticians. The two key difficulties are common to other similar contexts: (1) developing a website that attracts sufficient initial interest and (2) ensuring an honest system of ratings that provide reliable information about who is a good collaborator.

(12/27/12) Flexible software for statistical modeling: Perry de Valpine, Duncan Temple Lang, Daniel Turek, and I are developing hierarchical modeling software called NIMBLE. The aim is to enable one to specify a hierarchical model in BUGS-like syntax and then fit the model via a variety of different algorithms (different flavors of MCMC, sequential MC, maximum likelihood, etc.). The key idea is to separate model structure from fitting algorithm and make it easy to try different fitting approaches.

(12/27/12) Swapping corporate taxes for personal income taxes on stockholders This wouldn't fly in the current political environment, but how about we reduce corporate taxes to zero and increase personal taxes on the income brackets that tend to own corporate stocks to make up the difference. The basic idea (as seen with corporations playing tax avoidance games) is that corporate taxes are easier to game than personal taxes and result in long-term offshoring of corporate profits.

I'll close with a shout-out to Victoria Stodden, whose ideas diary partially sparked the idea for this page.

© 2012 Christopher Paciorek
Template design by Andreas Viklund / Icons from / Best hosted at