lost in translation...
Section III: Reporters Don't Lie - They're Just Stupid
in my opinion, the most problematic realm of scientific research has nothing to do with science, statistics, researchers, or their human slaves: it's that fuzzy translation medium that exists between published results and the public at large. i'm talking about the media, politicians (or more likely, their human slaves), and other "lay-type" folks, including doctors [yes, scary] - in other words, people that read about science but don't really understand it, but know enough techy-speak to convince their victims, excuse me, i mean the public, that they in fact know what they're talking about.
"what the hell does that mean?" there are actually three measures of central tendency: mean, median, and mode. most people don't know what the fuck a "mean" is, but they know "average" (a mean is in fact the arithmetic average). median is simply the middle score in a distribution, while mode is the most frequent score. many distributions are multimodal. many are not uniform, but skewed (read as "not bell-shaped"). why the fuck am i rambling about this? because when a reporter, politician, doctor, priest, or some other jerk-off tells you that the "average" is so-and-so, you tend to assume they mean the mean. but you don't know, and it's quite possible that neither do they, because statisticians are in the employ of people with biases as well, and they need to please their masters too. "average" could refer to any one of the three measures of central tendency. furthermore, this lovely number tells you nothing about the distribution of scores within a sample - what's the standard deviation (how spread out is it)? is it skewed?
am i being anal about this? you bet i am, but it's important, damnit! if you tell me americans are doing fantastic because the average income is $XK, but as it turns out the mode is $XK-$20K, that's pretty misleading. and they do shit like this all the time.
it's a fact, jack. this is actually one of my pet peeves. a significant result is never a fact. data points are facts. things you can measure - number of eyeblinks, height, surface area - these are facts. "people like food" is not a fact - it is a conjecture that is strongly supported by the facts. similarly, scientists never "prove" anything. their research provides support for ideas, theories, etc. nobody "proved" that taking aspirin lowers your chance of heart attack, and they never will. all they did was provide evidence that suggested or strongly supported the idea that it is true.
the aspirin research example provides another tender morsel: correlation does not equal causation. all that study showed was a correlation between taking aspiring and reduced likelihood of hear disease. the two are significantly related - they occur together - and the relationship happens to be inverse - when one goes up, the other goes down. this does not mean that taking aspirin causes a lowering of heart disease. in order to "provide support" for that idea you would have to do an experimental manipulation. yet the result is reported in the press as "taking apirin lowers [causation] your chance of heart attack". it may be that some unknown factor is responsible for the relationship.
"your biases were biased, and it's just like you, just like you..." politicians are biased, obviously. so are the media. so is your doctor. scientific results almost never come through this filter undistorted. un-replicated studies get celebrated as new "proof". conflicting studies get ignored or even ridiculed. any caveats expressed by the researchers concerning a result get swept under the rug.
and then, and then... by the time the "facts" get through this morass to the public at large, what you are left with is a stripped down, simplified, twisted, frankenbeast of an idea. and even that gets screwed up. how many times have you heard someone talk about something they heard on the television, only they got it all wrong, or worse yet, jumped to the completely wrong-ass conclusions because of their own biases.
even the best-case scenario is still horrible: they bought it, as is - the whole stinking mess...
in my opinion, the most problematic realm of scientific research has nothing to do with science, statistics, researchers, or their human slaves: it's that fuzzy translation medium that exists between published results and the public at large. i'm talking about the media, politicians (or more likely, their human slaves), and other "lay-type" folks, including doctors [yes, scary] - in other words, people that read about science but don't really understand it, but know enough techy-speak to convince their victims, excuse me, i mean the public, that they in fact know what they're talking about.
"what the hell does that mean?" there are actually three measures of central tendency: mean, median, and mode. most people don't know what the fuck a "mean" is, but they know "average" (a mean is in fact the arithmetic average). median is simply the middle score in a distribution, while mode is the most frequent score. many distributions are multimodal. many are not uniform, but skewed (read as "not bell-shaped"). why the fuck am i rambling about this? because when a reporter, politician, doctor, priest, or some other jerk-off tells you that the "average" is so-and-so, you tend to assume they mean the mean. but you don't know, and it's quite possible that neither do they, because statisticians are in the employ of people with biases as well, and they need to please their masters too. "average" could refer to any one of the three measures of central tendency. furthermore, this lovely number tells you nothing about the distribution of scores within a sample - what's the standard deviation (how spread out is it)? is it skewed?
am i being anal about this? you bet i am, but it's important, damnit! if you tell me americans are doing fantastic because the average income is $XK, but as it turns out the mode is $XK-$20K, that's pretty misleading. and they do shit like this all the time.
it's a fact, jack. this is actually one of my pet peeves. a significant result is never a fact. data points are facts. things you can measure - number of eyeblinks, height, surface area - these are facts. "people like food" is not a fact - it is a conjecture that is strongly supported by the facts. similarly, scientists never "prove" anything. their research provides support for ideas, theories, etc. nobody "proved" that taking aspirin lowers your chance of heart attack, and they never will. all they did was provide evidence that suggested or strongly supported the idea that it is true.
the aspirin research example provides another tender morsel: correlation does not equal causation. all that study showed was a correlation between taking aspiring and reduced likelihood of hear disease. the two are significantly related - they occur together - and the relationship happens to be inverse - when one goes up, the other goes down. this does not mean that taking aspirin causes a lowering of heart disease. in order to "provide support" for that idea you would have to do an experimental manipulation. yet the result is reported in the press as "taking apirin lowers [causation] your chance of heart attack". it may be that some unknown factor is responsible for the relationship.
"your biases were biased, and it's just like you, just like you..." politicians are biased, obviously. so are the media. so is your doctor. scientific results almost never come through this filter undistorted. un-replicated studies get celebrated as new "proof". conflicting studies get ignored or even ridiculed. any caveats expressed by the researchers concerning a result get swept under the rug.
and then, and then... by the time the "facts" get through this morass to the public at large, what you are left with is a stripped down, simplified, twisted, frankenbeast of an idea. and even that gets screwed up. how many times have you heard someone talk about something they heard on the television, only they got it all wrong, or worse yet, jumped to the completely wrong-ass conclusions because of their own biases.
even the best-case scenario is still horrible: they bought it, as is - the whole stinking mess...