Quote:
Originally Posted by PaleMelanesian
There is a definite leveling off of any gains past about 80 degrees. I don't use AC, so that's not the reason. I'm guessing 80-90 is the optimum intake temperature for my car.
I have very little data at the cold end, as you can see, so that's somewhat suspect. Most of this winter's points were on the high end there, while last winter had the low points. I think early results skewed the data. That, or just unfamiliarity with winter driving for mileage.
|
Pale,
Wow, lots of data points - what a nice change. However, don't you thinkk the distribution of those data points indicate that huge variables other than temperature are effecting your mpg? I'm not sure it is possible to pull the temperature component out like you have done. Your r* value of .16 tends to bear this out by indicating a very low correlation between the x and y axes. (For those who may not know, an r* (r-squared) value of 1 = perfect correlation, while 0 = no correlation).
For example, at 71F your data shows a high of 80 mpg and a low of 35 mpg (and everything in between). That's a difference of 45 mpg! Clearly some unknown variables are having a very large effect on your mpg even when temperature is constant. Therefore, the smoothed regression line calculated from data with an r* of 0.16 has to be taken with a huge grain of salt. A problem with Excel charts is that they often imply stronger correlation than is actually there because Excel will draw a nice regression line even when one may not really be appropriate.
Another problem in accepting that this line accurately represents temperature's effect on mpg is that seasonal variables independent of temperature are not controlled for. For example, is some of your winter commute in the dark when the days are shorter? Traffic generally moves more slowly in the dark than in daylight and would reduce your mpgs. Or does the lower winter Sun shine in driver's eyes, slowing traffic? Neither of these confounding variables would have anything to do with ambient temperature
per se, yet would produce an apparent correlation when T and mpg are plotted against each other on a graph. This is the bugaboo that catches so many researchers - their study actually measures an unrelated correlation rather than the cause and effect they were thinking about.