But I want to know how can I analysis a probability plot like that? I want the way of analysis the probability plot.
If data is normally distributed the points in the probability plot all appear near or on the ideal line (blue). The points should be linear, mainly in the middle of the values measured. A rough test for normality with probability plots is called "fat pencil test": If (nearly) all points disappear if you put a fat pencil on the ideal line, the data can be considered to be normally distributed.
When stacked points appear (like in your plot) identical values among the measurements are present in the data. A common reason for the stacks is a small resolution of your gage.
The normal distribution is a continous distribution, so there is (theoretically) only a very small probability to get identical values within 125 values total. If the process outcome is in fact normally distributed but the resolution of your gage is too small to get the differences between the parts measured, you will alway reject the assumption of a normal distribution. Therefore it is required to have a gage with appropriate resolution for the interval measured to get reliable results out of the probability plot and test for normality.
Compare the individual value plots for your measurements and random normal data as well as the probability plots for both attached and you will get a clearer picture what an ideal state is and why the assumption of normality in your data is rejected.
Regards,
Barbara