0

enter image description here

I wanted to investigate if my data is normally distributed with a QQ Plot. I'm not quite sure if the deviations from the theoretical plot are too big to crate a bland-altman plot?

C K
  • 51
  • 2
    "Too big" for what? What do you need normality for? – Christian Hennig Jul 08 '22 at 15:20
  • 1
    to create a bland-altman plot – C K Jul 08 '22 at 16:10
  • 1
    As far as I can see, Bland & Altman simply take a Tukey mean-difference plot and stick mean $\pm$ 2 sd lines on it. There doesn't seem to be a strong need to assume normality to make the plot; the only thing that might be at all in question is the suitability of the additional lines, but even that's not necessarily a huge issue (e.g. where they talk about the proportion of points within 2 sd of the mean that's not so sensitive to normality). 1. What are you using the plot to do, exactly? 2. What's the cause of the spike of values near -1 in your data? – Glen_b Jul 09 '22 at 04:05
  • But doesn't the difference of the data have to be normally distributed to be able to calculate the SD's and the 2.5% and 97.5% percentiles? – C K Jul 09 '22 at 11:42
  • The QQ plot is only used to see if the differences of the data are normally distributed – C K Jul 09 '22 at 11:43

1 Answers1

1

I mean, in the tails you're seeing two sample standard deviations map to 2.5 and 3 normal standard deviations. That's a red flag if your concerned about tail probabilities.

That said, does Bland--Altman plots really require normality? If you use them to compute confidence intervals, sure, but not for plotting.

kqr
  • 814