It’s true that the downward trend can’t go on forever, and to say that it’s definitely going to continue would be (all by itself, without some other arguments) an appeal to history or slippery slope fallacy. However, when we see a trend as consistent and as potentially meaningful as the one below, it makes sense to start wondering why it is happening:
I was mostly just trying to point out that you are extrapolating from a sample size of three points. Three points which have a tremendous amount of common causes that could explain the variation. Furthermore you aren’t extrapolating 10% further from the span of your data, which might be ok, but actually 100% further. You’re extrapolating for as long as we have data, which is… absurd.
One, I am used to seeing the term “sample size” applied to things like the people being studied, not a number of points done in a calculation. If there is some valid use of the term “sample size” that I am not aware of would you mind pointing me in the correct direction?
Two, I am not sure where you’re getting “three points” from. If you mean the amount of IQ points that LessWrong has lost on the studies, then it was 7.18 points, not three.
Three points which have a tremendous amount of common causes that could explain the variation.
Two points per year, which could be explained in other ways, sure. No matter what the trend, it could be explained in other ways. Even if it was ten points per year we could still say something like “The smartest people got bored taking the same survey over and over and stopped.” There are always multiple ways to explain data. That possibility of other explanations does not rule out the potential that LessWrong is losing intelligent people.
Furthermore you aren’t extrapolating 10% further from the span of your data, which might be ok, but actually 100% further.
Not sure what these 10% and 100% figures correspond to. If I am to understand why you said that, you will have to be specific about what you mean.
You’re extrapolating for as long as we have data, which is… absurd.
Including all of the data rather than just a piece of the data is bad why?
Three points referred to the number of surveys taken, which I didn’t bother to look up, but I believe is three.
10% and 100% referred to the time span over which these data points referred to, ie. three years. Basically, I might be OK with you making a prediction for the next three months (still probably not) but extrapolating for three years based on three years of data seems a bit much to me.
Oh I see. The problem here is that “if the trend continues” is not a prediction. “I predict the trend will continue” would be a prediction. Please read more carefully the next time. You confused me quite a bit.
If you’re not making a prediction, then it’s about as helpful as saying “If the moon crashes into North America next year, LW communities will largely cease to exist.”
It’s true that the downward trend can’t go on forever, and to say that it’s definitely going to continue would be (all by itself, without some other arguments) an appeal to history or slippery slope fallacy. However, when we see a trend as consistent and as potentially meaningful as the one below, it makes sense to start wondering why it is happening:
IQ Trend Analysis
I was mostly just trying to point out that you are extrapolating from a sample size of three points. Three points which have a tremendous amount of common causes that could explain the variation. Furthermore you aren’t extrapolating 10% further from the span of your data, which might be ok, but actually 100% further. You’re extrapolating for as long as we have data, which is… absurd.
One, I am used to seeing the term “sample size” applied to things like the people being studied, not a number of points done in a calculation. If there is some valid use of the term “sample size” that I am not aware of would you mind pointing me in the correct direction?
Two, I am not sure where you’re getting “three points” from. If you mean the amount of IQ points that LessWrong has lost on the studies, then it was 7.18 points, not three.
Two points per year, which could be explained in other ways, sure. No matter what the trend, it could be explained in other ways. Even if it was ten points per year we could still say something like “The smartest people got bored taking the same survey over and over and stopped.” There are always multiple ways to explain data. That possibility of other explanations does not rule out the potential that LessWrong is losing intelligent people.
Not sure what these 10% and 100% figures correspond to. If I am to understand why you said that, you will have to be specific about what you mean.
Including all of the data rather than just a piece of the data is bad why?
Three points referred to the number of surveys taken, which I didn’t bother to look up, but I believe is three.
10% and 100% referred to the time span over which these data points referred to, ie. three years. Basically, I might be OK with you making a prediction for the next three months (still probably not) but extrapolating for three years based on three years of data seems a bit much to me.
Oh I see. The problem here is that “if the trend continues” is not a prediction. “I predict the trend will continue” would be a prediction. Please read more carefully the next time. You confused me quite a bit.
If you’re not making a prediction, then it’s about as helpful as saying “If the moon crashes into North America next year, LW communities will largely cease to exist.”