 |
Re: Report: 3 Amtrak workers in fatal 2016 crash had used drugs
Author: BOB2
Date: 01-28-2017 - 11:46
Interesting statistic "43%" increase in positives from the 50,000 random tests....so what exactly does that mean if you actually passed statistics?
What percentage of the 50,000 actually tested positive?
How many actual total positive tests was that?
Because, if say 100 of the 50,000 tested positive last year, that would be .002%, and this year, if 143 out of 50,000 tested positive, that would indeed be a "43%" increase in positive tests. But, the total number testing positive would only rise to .00286%......of those 50,000 tested.
So how is the sampling variation to be expected based on this and previous testing years? And, what was the "bound of error" for the likely variation of the samples around the "true" mean change to determine rejection of the null hypothesis, that there was really no significant statistical change in the actual "mean" level of drug use in the population as a whole, as opposed to the tiny change in the sampling?
Since, reporting that this may be "white noise" (variation around the actual mean from multiple samples) doesn't make for as good of "story", the editors clearly needed a "bigger" number for that dramatic effect. And, doesn't the editor have a duty to pander to the truly ignorant with a "big" headline "number", and after all, 43% is a much bigger "number" than .00286%......, right? Wrong.....!
Of course, without know the base sampling numbers from which that percentage was calculated, and knowledge of the bound of error, based on this and previous sampling, it is meaningless bull$hit, for making any meaningful statistical inference as to this purported change in the overall level of purported drug use in the RR industry.
There have been tragic consequences of drug and alcohol impairment in the RR industry and it is an important subject. That selective number, which itself may have no meaning at all, used to "sex up" (as the Brits like to say) a story, does this serious issue a disservice.
But. based upon some of these goofier responses this post has elicited, it does appear, that for agitprop (old fashioned Soviet style "agitation and propaganda") purposes, such dubious statistical nonsense appears to work quite well though.
What percentage of the 50,000 actually tested positive?
How many actual total positive tests was that?
Because, if say 100 of the 50,000 tested positive last year, that would be .002%, and this year, if 143 out of 50,000 tested positive, that would indeed be a "43%" increase in positive tests. But, the total number testing positive would only rise to .00286%......of those 50,000 tested.
So how is the sampling variation to be expected based on this and previous testing years? And, what was the "bound of error" for the likely variation of the samples around the "true" mean change to determine rejection of the null hypothesis, that there was really no significant statistical change in the actual "mean" level of drug use in the population as a whole, as opposed to the tiny change in the sampling?
Since, reporting that this may be "white noise" (variation around the actual mean from multiple samples) doesn't make for as good of "story", the editors clearly needed a "bigger" number for that dramatic effect. And, doesn't the editor have a duty to pander to the truly ignorant with a "big" headline "number", and after all, 43% is a much bigger "number" than .00286%......, right? Wrong.....!
Of course, without know the base sampling numbers from which that percentage was calculated, and knowledge of the bound of error, based on this and previous sampling, it is meaningless bull$hit, for making any meaningful statistical inference as to this purported change in the overall level of purported drug use in the RR industry.
There have been tragic consequences of drug and alcohol impairment in the RR industry and it is an important subject. That selective number, which itself may have no meaning at all, used to "sex up" (as the Brits like to say) a story, does this serious issue a disservice.
But. based upon some of these goofier responses this post has elicited, it does appear, that for agitprop (old fashioned Soviet style "agitation and propaganda") purposes, such dubious statistical nonsense appears to work quite well though.
