@tjreese I never wanted this part of the thread gone, just the title changed - this is a very interesting debate, imo. I've been on both sides of it in my life... turned 40 last month, so I'm probably in the median range age here. I value and see validity in both points of view. I've reread the thread and appreciate the time and thought given to it by everyone. I'll keep reading it... and hoping for tv times all the same
tic.
Let me try to summarize. It is my belief that proudpurdue considers global warming, global cooling or just climate change to be settled science. That is fine, he may be right in the projected results, but science is far from proven. I think he and others in his group read circumstantial evidence and believe they have enough evidence to support their stance. When dealing with a statistic or sampling data in trying to describe the population it is important that the data represent the population and so becomes the question what you are trying to project and does your data cover enough of the area to make that prediction. No matter the particular shortcomings of sampling, you need your measured response to be repeatable and reproducible as well as accurate. Essentially, sampling errors can lead to many false conclusions. Measurement errors can mask or hide the potential differences or significance in addition to mean shifts. In other words, not sampling over the population being reference opens the data to many misinterpretations...as would any heavy population pulls weighting the data and/or biased sampling. The measurement error side…even if accurate, but not repeatable or reproducible provides excessive variation under the same event and hides what could be a statistically significant result and that may or may not be a significant result in real life.
IN other words, statistical significance in any study says the variable (s) in question shows an unlikely difference in means due to chance. In a straight hypothesis study an F-test is used to determine if the populations of the two means have similar variations and if not the degrees of freedom are adjusted to compensate for this. Bottom line…is the difference in means far enough apart when the variations are considered that it could have difference values and still not be significant. THAT concept is VERY important.
Merely seeing different values is meaningless. In ANOVA’s (more than one condition is studied) variation may be assumed to be equal (studying tire wear on the frt driver’s side under different levels or pressure or tire manufacturers etc.) Regression is that ANOVA type study over time…over some change where we look for slope change.
Now, let’s go back to that tire study. Let’s assume that I only care about a certain tire size and maybe only two manufacturers of tires relative to the wear on that tire under various tire pressures. Let’s say rubber compound, expected load range and such suggests a nominal 35 PSI for test. Let’s say my tire gage on a perfectly flat surface provides “some” level of repeatability and reproducibility for a reasonable anticipated study and my pressure adjustments always take place on this surface for representative inflation for study…and input variable. Furthermore, let’s pretend we also have a very repeatable tread wear instrument and throw out that variable as well…both that could cause false results.
We will also pretend we are smart enough to know that not only car, but tire location on every car would provide different results and so we decide to only study the frt tire on the driver’s side AND on a certain car model…fully understanding that the results and conclusions may not be accurate for other locations or car models not studied…we are that smart. How do we know for certain what happens if not studied? Again, this is a pretty simple study with no inflation or tire wear error to complicate this. We decide to run this tire comparison for wear by inflating each tire studied at 32,33,34,
35,36,37 and 38 PSI. I can manipulate the variables else I always struggle with any correlation data being causal or not. I choose the very middle of the tread for my tire wear measurements I take 7 tires from each manufacturer and inflate them to each level of pressure and compare the initial tread depth to the depth after some specified mileage or mileages. My tests are on the same surface, moisture and temperature as well as an automated control of my speed to test and consistent speed across test. I measure the wear under as controlled conditions as possible under the initial PSI, a controlled environment and surface for the specified mileage and re-measure with my perfect instrument to determine the wear from my initial reading absent any error to each endpoint absent any error in controlled conditions affecting my numbers registered in the middle of the tire.
The results may or may not have a linear relationship between pressures and wear mileages. It is quite possible that wear takes place more or even less with new tires. Do the new tires exist with the same depth initially? Does the tread depth initially have a different result due to harness differences through the depth of the tread? How is the variation of each manufacturers tire models change in hardness, tread depth consistency, sidewall rigidity and any other variable that may come into effect? If a variable in the manufacturer of the tire effects wear and we have no repeated measures under each condition can we draw conclusion that may not be correct? If we only measure the middle of the tire, is that wear reflected accurately when the tire has less air or more air…is the wear linear or does the variation of wear take a different path by not measuring the edges of when a tire is underinflated but not studied. Can I say the wear is the same whether it is in all four locations? Can I say the wear is the truth and assume that holds for all cars…for all locations in a car…for all driving conditions, for degrees of toe in or out…so very many things that “might” draw a different conclusion and all of that with no error and a perfect measuring system, input system and controllable environment? Do I conclude that tire hardness was the variable that effected the resulting data, not knowing the tread design or sidewall rigidity?
There are many more questions that arise when the data shows a certain wear…things as simple as …does the wear under these same conditions continue after 70,000 miles (not studied) as it did in the first 20,000 miles? Do we assume that is a linear relationship that extends well past the data? Did I randomize my testing or did anything in the car show up that confounded the data I was unaware…looser alignment, tie rod slop increase, inappropriate lug tightening for perpendicularity etc…etc…
There are a lot of professional people in this forum and I’m guessing that these professional people may or may not have considered this very small, simple, finite study being quite as complicated as it is. If questions along this line are not answered it would not matter if 99.9% of a group found indicators or circumstantial evidence…it is still not proven…and may never be proven…and certainly no confirming experiment exists to recreate the conditions for proof…it is an educated hunch…and “MAY” be correct, but a hunch that “should” generate further study before conclusions are drawn or the results are carefully worded to not depict something that is unknown.
In summary I can show without question that foot size is highly correlated to academic achievement...much more than any other variable for both males and females within a certain range of foot sizes. I could wager all I own against others and I would be right…in spite of foot size increases not creating higher academic achievement even though the academic data slopes upward as the foot size increases. Yeah, I still have some questions even though others may be convinced.