RE: Apples, Damn Apples and Statistics - Data Reading
I'm amazed what kinds of posts I stumble upon on this platform from time to time. Really dig it, man! Hardcore info is my piece of shit. (no pun intended in relation to your last image)
Now, with the Gould example, I gather that his doctor didn't know how to interpret the median. If statistics knowledge isn't required for doctors (I don't know whether it is), then why don't they give a 95% confidence interval in medical studies, rather than a median?
I imagine there is a way to do a 95% confidence interval, although I've only seen it with bigger sample sizes (usually recommended is above 30). Even with a right-skewed distribution like that, you can calculate the means of the samples and then create the sampling distribution, which is always a normal distribution. So you can calculate a 95% confidence interval. But do you know if it can be calculated if the sample size is 1?
Also, from TV there is ingrained the notion that the doctor tells you you have so much time left to live. Rather than that he is reasonably confident there is this much chance you will live at minimum this much and at maximum that much time. (note the expression: reasonably confident, this much chance, time interval)
No, you can't interpret that from what I wrote. If you check again I didn't say his doctor told him he had 8 months to live. He read literature on his own and his heart most likely skipped a beat when he read the median. The whole thing probably didn't last more than a couple seconds and that inspired his article.
If you check the first reference.
He didn't mention what prognosis his doctor gave him. Just that she didn't give him literature to read.
Most medical studies have a confidence interval and statistical significance in them. Even at the time of his diagnosis 1982, is fairly probable his doctors had to use C.I.
Most likely what happened, since I can't find another recorded mention of what happened with his doctor, she just underestimated him. Even if she had given him a risk profile analysis, like the one he did on its own, that is not a guarantee of anything. We can now see the right thing, but that's mostly survivorship bias.
Indeed, the medical training on statistics is lacking, but that's true of most fields where their participants are not actively involved in research. Since the apparition and popularization of Evidence-Based Medicine (EBM) Giving such information to patients is routine and most of the population is actually more troubled by it than relieved.
OK, thanks for clearing up my misinterpretation.
No problem :)