I was thinking the same thing, ie, MBA's aren't required to have calculus, so why use a statistical function (linear regression; heavy with calculus) for this analysis? What will you find out from it?

Psychologists do the same thing, using statistical function they don't understand and that may not even apply to the problem.

From another Engineer who is not even funny.

 

I agree regression is WAY overused to draw assumptions of causation where correlations just happen to appear (like skirt length and stock market returns were highly correlated for like 60 years, literally).

Especially when modeling more complex systems like weather, climate, stock market returns, biological systems, etc. Where chaos exists, regression is a bad model because it implies linear, fixed relationships where there is really a more dynamic independent/independent/dependent relationship.

However, you'd be hard pressed to make an argument that a race track is totally chaotic. Power to weight and sticky tires are ALWAYS going to be very highly positively correlated with lap times. You don't need to regress anything to figure that out.

Are you proposing that this isn't the case? It seems like one of the few areas where some decent math might actually apply (as opposed to assuming that a certain correlation between historical interest rates and stock market returns will continue, it won't!)

The only reason this is interesting is because it demonstrates just how strong the relationship is between power/weight and lap time and just how good the engineers are with modern suspensions (as they've removed a lot of the variance associated with this independent variable). That's it.

 

I find this comment interesting:

"of the 9 cars faster than 1 SERY, 6 were Porsches"

This could mean Porsche has a better driver than most of the other manufacturers, they run their laps on days when the weather/temperature is more favorable, of MAYBE as many of us on this site feel, the overall design of Porsche vehicles (suspension, brakes, weight distribution,...) is better than most other vehicles on the road. Maybe Porsche really can get more performance from their weight/hp ratio than others?

 

Maybe Porsche claimed hp numbers are more accurate than others?

Marton

 

I think it's probably some combination of 40 years of engineering and racing this platform, the nurburgring being close by and relatively lower weight (independent of power to weight), but the cynical almost engineer in me says the biggest factor is probably walter rohrl driving...

 

Derrick,

If you want to be geeky and see if this fits, then calculate the R^2 for your model, not just what percentage falls into a range. That will be a good measure of whether the model makes sense or not.

 

I should have included that, sorry. R^2 for the base model is about 0.73, but it jumps to about 0.79 when you isolate most of the data from track tires/suspension, heavier cars and lower powered cars.

So for an "average" sports car of around 1400kg and 350hp (without track tires), roughly 80% of variance in track times can be explained solely by p/w ratio.

 

x2. Although all other manufacturers also have drivers comparable to Walter Rohrl driving their cars. His familiarity with the track, and Porsche's receptivity to his tuning inputs (specifically at the Nurburgring) should play a significant part of the error.

Another interesting variable to consider would be the country of the manufacturer. All else equal, I would imagine cars to run faster at their home track (Porsche at Nurburgring, Honda at Suzuka, etc.).

Thanks for the interesting analysis.

 

All Formula 1 cars in England

Marton

 

Can you whip up a quick regression model for my mother-in-law's annoyance to weight ratio? I'm not sure if the results will be linear, more likely parabolic. An interesting comparison would contrast that to an annoyance to (weight + alcohol) ratio.

Thanks.

-td

Oh, thanks for the original post, it was quite illuminating!

 

I already did it for my mother in law. It has nothing to do with weight but is 100% correlated with proximity. If you throw alcohol into the equation (for me, not her) the correlations went WAY over 100%, so much so that the only way to solve the equation is to quickly and radically reduce the proximity factor.

Unfortunately, this is apparently negatively correlated with other factors having to do with the daughter of the mother in law.