I Help me understand skewness in QQ-plots please

  • I
  • Thread starter Thread starter bremenfallturm
  • Start date Start date
bremenfallturm
Messages
81
Reaction score
13
TL;DR Summary
I am trying to understand how QQ plots work, but I have a hard time understanding how to interpret skewness. Specifically, it is "the other way around" than I expect. See the post for an explanation.
I am trying to understand how QQ plots work, but I have a hard time understanding how to interpret skewness. Specifically, it is "the other way around" than I expect.

Let me explain.

From what I understand, in a QQ plot, we divide the normal distribution (typically ##N(0,1)##) and the dataset into ##n## quantiles (where ##n## is the number of datapoints). We sort the dataset and plot each datapoint against the normal distribution. For example, if we have 10 ordered datapoints $$a_1, a_2, ...$$, and have created 10 normal quantiles $$n_1,n_2,...$$ we would plot $$(a_1, n_1), (a_2, n_2)$$ and so on.

Now, here is when I don't understand how we interpret the skewness.
Consider the left skewed case for example (https://anasrana.github.io/ems-practicals/qq-plot.html)
1746684655365.webp

If I look at the plot, my first intuition is this: it looks to me that all points below the line (the points between -4 and around -1 of the normal distribution's quantiles) are smaller than expected. This is because they are below the line. Therefore, the points would be drawn from a distribution where smaller values are more probable. Of course, looking at the actual distribution, we can see that it is the other way around.
My second idea is then this: if we have many large datapoints (i.e. in the image above), the graph axes are going to be scaled such that the smaller values fall below the line, and thus, we have a distribution with a tendency towards large datapoints. Does any of this make sense? Could you help me deepen my understanding?
 
Physics news on Phys.org
bremenfallturm said:
If I look at the plot, my first intuition is this: it looks to me that all points below the line (the points between -4 and around -1 of the normal distribution's quantiles) are smaller than expected. This is because they are below the line. Therefore, the points would be drawn from a distribution where smaller values are more probable.
True.
bremenfallturm said:
Of course, looking at the actual distribution, we can see that it is the other way around.
No it isn't. It's exactly as expected.
bremenfallturm said:
My second idea is then this: if we have many large datapoints (i.e. in the image above), the graph axes are going to be scaled such that the smaller values fall below the line, and thus, we have a distribution with a tendency towards large datapoints. Does any of this make sense?
No, none of this makes sense.
 
Hi all, I've been a roulette player for more than 10 years (although I took time off here and there) and it's only now that I'm trying to understand the physics of the game. Basically my strategy in roulette is to divide the wheel roughly into two halves (let's call them A and B). My theory is that in roulette there will invariably be variance. In other words, if A comes up 5 times in a row, B will be due to come up soon. However I have been proven wrong many times, and I have seen some...
Thread 'Detail of Diagonalization Lemma'
The following is more or less taken from page 6 of C. Smorynski's "Self-Reference and Modal Logic". (Springer, 1985) (I couldn't get raised brackets to indicate codification (Gödel numbering), so I use a box. The overline is assigning a name. The detail I would like clarification on is in the second step in the last line, where we have an m-overlined, and we substitute the expression for m. Are we saying that the name of a coded term is the same as the coded term? Thanks in advance.
Back
Top