Using this plot as a statistical argument against the idea of Dunning Kruger is pretty Dunning Kruger.
It’s a scatter plot with hella dense overlap and no density indicator, and a best-fit line that doesn’t match the visual skew. Different ranges and scales on the axes are wonky as hell and make it hard to follow a 1:1 reference line. Try tracing a mental line from (75,75) to (150,150) and it looks much cleaner.
It seems to show subjective vs. evaluated IQ results, which mostly shows people generally think they’re far above average. There’s clearly a statistically significant positive correlation in the trend of modest magnitude.
More importantly though, it’s a bit of a weird thing to tie to Dunning Kruger: The plot tells us something about how well people can guess how “cognitively gifted” they are relative to peers, but that’s not a reasonable proxy for how people might self-evaluate on a skill-based or knowledge-based domain.
I’ll correct myself, the data is simulated and I skipped the word. You are completely right to remind me of this. But there are a few important considerations here.
The original DK study used a IQ like score based on general cognitive faculties for its argument, subjective vs. measured.
Wiki quote: “The initial study by David Dunning and Justin Kruger examined the performance and self-assessment of undergraduate students in inductive, deductive, and abductive logical reasoning; English grammar; and appreciation of humor.”
Second of all, for this I have consulted the study itself, according to my cursory read the sample size of the study seems to be below 100. I’ll let that speak for itself.
A new study working with more data, that doesn’t just present generalised skill shows that, just like in the simulated data, it is still a blob following a normal distribution in terms of density. The simulated data was meant to show how misleading inferences can be made with noisy results.
Using this plot as a statistical argument against the idea of Dunning Kruger is pretty Dunning Kruger.
It’s a scatter plot with hella dense overlap and no density indicator, and a best-fit line that doesn’t match the visual skew. Different ranges and scales on the axes are wonky as hell and make it hard to follow a 1:1 reference line. Try tracing a mental line from (75,75) to (150,150) and it looks much cleaner.
It seems to show subjective vs. evaluated IQ results, which mostly shows people generally think they’re far above average. There’s clearly a statistically significant positive correlation in the trend of modest magnitude.
More importantly though, it’s a bit of a weird thing to tie to Dunning Kruger: The plot tells us something about how well people can guess how “cognitively gifted” they are relative to peers, but that’s not a reasonable proxy for how people might self-evaluate on a skill-based or knowledge-based domain.
E: and I went looking for the source and that’s fake, simulated data you’re presenting as real there lol https://en.wikipedia.org/wiki/Dunning–Kruger_effect#Statistical_and_better-than-average_effect
I’ll correct myself, the data is simulated and I skipped the word. You are completely right to remind me of this. But there are a few important considerations here.
Wiki quote: “The initial study by David Dunning and Justin Kruger examined the performance and self-assessment of undergraduate students in inductive, deductive, and abductive logical reasoning; English grammar; and appreciation of humor.”
Second of all, for this I have consulted the study itself, according to my cursory read the sample size of the study seems to be below 100. I’ll let that speak for itself.
A new study working with more data, that doesn’t just present generalised skill shows that, just like in the simulated data, it is still a blob following a normal distribution in terms of density. The simulated data was meant to show how misleading inferences can be made with noisy results.