Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 2 down vote favorite
1

First, I apologize for the bad image, but it's the best I can do with a mouse. Anyways, here it is:

enter image description here

So the data is skewed right, but the normal probability plot bends up and over what would be the approximate linear equation. According to different websites1, 2, a normal probability plot with data skewed right goes under the approximate line of best fit, and my graph looks more like it's skewed left. What am I misunderstanding?

1: itl.nist.gov/div898/handbook/eda/section3/normprp4.htm
2: www.basic.northwestern.edu/statguidefiles/probplots.html#Data%20Skewed%20to%20Right

share|improve this question
The graph is impossible to understand without labels and scales for the axes. – whuber♦ Oct 5 '11 at 3:13
It's a normal probability plot from -3 to 3 on the y-axis and time on the x axis. – Anonymous Oct 5 '11 at 3:14
Assuming time increases left to right and y increases bottom to top and that the normal plotting positions have been correctly computed, this is a plot of right-skewed times. Why does it look skewed left to you? – whuber♦ Oct 5 '11 at 3:18
The websites show that right skewed graphs don't look like that, while it looks like the examples for left skewed graphs. – Anonymous Oct 5 '11 at 3:21
1  
Is that a frequency histogram at the top? If you best want to understand your question, try making the normal probability plot yourself, without a computer. – Adam Oct 5 '11 at 3:55

1 Answer

up vote 8 down vote accepted

This graph interchanges the axes compared to the cited websites, that's all.

In general, to read a probability plot, ask yourself "what changes in the data would be required to line the points up diagonally?" In this case, the data are shown on the x-axis, so changing the data would slide points horizontally along the x-axis while retaining their vertical positions. To get the points into a diagonal line we would have to slide the largest (rightmost) times to the left (that is, pull them in towards their middle) and we would have to slide the smallest (leftmost) times a little to the left as well (that is, push them away from the middle time value). That tells us the large times are too big compared to a normal distribution: they are skewed towards large values (considered the "right," no matter how the plot is drawn; better terminology is "positively skewed"). (For reading q-q plots in general, I have posted a more elaborate explanation with illustrations.)

When the axes are reversed, the times (or, generally, the data) are plotted vertically and the sliding has to happen in the vertical direction. There's no chance of confusion, though--provided the axes are clearly labeled!

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.