Meaning of error bars in JASP's QQ plot?

AceOfBayes

Hi all,

I'm analyzing reaction time data in JASP. JASP makes it very convenient to check my data for normality by providing the QQ plot against a normal distribution. I noticed that the plotted data points all have error bars around them. This is in contrast to what I have seen in R's qqnorm function.

What do these error bars mean?

Can I use them to help me in this (sometimes quite subjective) decision whether or not my data are normally distributed?

Example: I have attached my QQ plots. On the left you can see that there is a deviation from normality at higher values. This is to be expected because RT data typically have longer tails towards the right. I took the natural logarithm of the RT and examined how that changed the QQ plot. If you look at the right-most data point indicated by the arrow, you can see that it has slightly moved closer too the line. The read line is now more comfortably within said error bars. Does this mean that the transformation has made the data more Gaussian? Are the untransformed data already "Gaussian enough" since the error bars already include the red line?

Cheers,

Michael

EJ

Hi Michael,

1. It is always somewhat of a judgment call
2. I don't find the change from RT to ln(RT) particularly striking
3. I was unable to get the error bars from JASP -- they do not show up for me in the Descriptives, nor in the Distribution module. Can you tell me exactly what you did to get them?

Cheers,

E.J.

AceOfBayes

Dear E.J.,

Thanks for your time!

1. True
2. All in all: neither did I :) I just wanted to provide an example where the meaning of these error bars may be heIpful to decide between the two versions. In the end, I decided to leave them as they are. Especially given that none of the Shapiro-Wilk's tests detected any normality violation. (Could be due to low power of course.)
3. That is strange. The Bayesian rm-ANOVA example from this manual also shows the same error bars. So you should be able to follow the example there to reproduce this behavior. I'm running the JASP version 0.13.1 on macOS High Sierra (10.13.6).

I was hoping that this might be some sort of highest density interval for believable values that the JASPers had kindly included to facilitate people's judgments.

Cheers,

Michael