The current "drug em till they don't ask tough questions and go outside the lesson plan" treatment for the fictional ADHD money-making diagnosis is all hogwash, lies, and downright child endagerment.(my opinion) A VERY small minority actually has something treatable by the snake oil concotions...
Walt On Nov 7, 2010 10:09 AM, "Peter Frederick" <psf...@earthlink.net> wrote: > Statistics are just numbers, and the formulas used to generate them > are NOT intelligent -- they work no matter what numbers you put in. > > The first lessons in a decent statistics class involve long > discussions of this fact -- to whit, the numbers mean nothing unless > the underlying assumptions of logical connection are true. The logic > and measurement parameters mean everything. > > One should also remember that descriptive statistics are just that -- > they describe a population one has measured. Bad measuring tools > result in bad data, at best, and descriptive statistics are NEVER > predictive. > > inferential statistics (things like comparing test vs control > subjects ) are predictive ONLY when all the assumptions made about > measurement and logic are true (watch for a term called "external > validity" -- it doesn't generate numbers, it's a logics problem). > This is where statistics tend to veer off into the land of fantasy, or > every outright lies. Just like a survey, one has to be careful about > what data is generated to apply statistics to, and how that data > actually applies to a problem, since if you don't measure what you > think you are, the results are meaningless. > > A current example is the use of amphetamines to treat ADD or ADHD -- > all the measurements are whether the kid sits still or doesn't disrupt > the classroom, so far as I know there are NO studies, even poor ones, > that indicate that the kids learn more. Bogus research, so far as I'm > concerned, and a gross misuse of statistical tools: the conclusions > are completely un-related to the measurements. Another bad result > here is the use of amphetamines by kids "studying" as the assume they > lead to better learning. No evidence of that, either. The logic > failure: compliance in the classroom is mis-labeled as "learning". > > Statistical process control is another thing that gets me fired up, > since it's almost always used improperly by American management > people. It's only useful if you use the data to control a device -- > can't be applied to employees, sales, or biological systems (with a > few exceptions), and even then is only useful if you have a calibrated > knob to turn. I've been told many times it's not useful for > production since it "only finds bad parts" -- typically, the person > telling me this missed the whole point, which is to use the > measurement data to fix the machines MAKING the parts so that only > good parts get made. A monkey with a go/no go gauge can find bad vs > good parts, no need of mathematics at all, but if you want to figure > out how to fix or replace a manufacturing tool BEFORE it makes off > sized stuff, SPC is a miracle. Totally worthless for a pH meter, > though -- no variation allowed in the measurements, if it's out of a > very narrow range, you re-calibrate or repair it. > > I could go on and on, but the point I'd like to pass on is that one > MUST know and understand the structure, measurements, and assumptions > made in any study of any type using statistics before reaching any > conclusions on the validity of the results. > > Peter > > _______________________________________ > http://www.okiebenz.com > For new and used parts go to www.okiebenz.com > To search list archives http://www.okiebenz.com/archive/ > > To Unsubscribe or change delivery options go to: > http://okiebenz.com/mailman/listinfo/mercedes_okiebenz.com _______________________________________ http://www.okiebenz.com For new and used parts go to www.okiebenz.com To search list archives http://www.okiebenz.com/archive/ To Unsubscribe or change delivery options go to: http://okiebenz.com/mailman/listinfo/mercedes_okiebenz.com
