Open Franck6S opened 3 weeks ago
Dear @Franck6S,
thanks for the bug report. The issue here is that the distribution module finds the parameters by maximizing the log likelihood function. The starting points of the optimization procedure are the parameter values set in the show distribution
section; by default it's shape: 1, scale: 1). The default is very far away from your data (whose scale is about 3 orders of magnitude larger than that of the default): the log likelihood function is too flat in that region and the optimization gets stuck at nonsensical values.
If I set the starting values to something more closer to the scale of your data:
I get much more sensible output:
I will check if we can detect such cases and give some informative message. In the meantime, I hope this helps!
JASP Version
0.18.3
Commit ID
No response
JASP Module
Distributions
What analysis are you seeing the problem on?
Weibull
What OS are you seeing the problem on?
Windows 10
Bug Description
While calculating weibull parameters, it gives wrong parameters for weibull distribution identification , please see related calculation with other software Please found related dataset : 1200 1400 1000 2280 1680 1828 1828 1310 2100 1100 950 1360 2170 1900 2120 100 1140 517 1550 1300 1380 600 1300 1800 1182 1500 1600 1800 1800 1450 900 1115 750 1441 1082 1270 1330 1800 580 1200 800 850 870 800 1100 1130 1240
Expected Behaviour
This should have similar results values as R or Minitab software
Steps to Reproduce
Log (if any)
Book5.xlsx
Final Checklist