Author Topic: Extreme cases and WLE  (Read 766 times)


  • Newbie
  • *
  • Posts: 14
    • View Profile
    • Email
Extreme cases and WLE
« on: November 13, 2016, 10:28:32 PM »
Hi folks,

I'm trying to compare the results produced by Conquest and Winsteps. So I ran the rating scale model with 123 samples on both Conquest (latest version) and Winsteps. When I plot the correlation of the person estimates (winsteps estimates on x-axis and Conquest estimates on y-axis), there seems not to be a perfect linear correlation. The r-squared is around 0.99, and a few extreme high and low scores  pull the linear relationship into a quadratic shape...I'm wondering if anyone has come across this situation before and explain why this happened?

Also, I'm using the WLE person estimates in Conquest to create the correlation plot, but I haven't found a clear explanation on the Conquest manual about the Weighted likelihood Estimation process, what is weighed, and how the weight is used. Has anyone had any idea?

Thank you very much for your help!


Eveline Gebhardt

  • Administrator
  • Full Member
  • *****
  • Posts: 103
    • View Profile
    • Email
Re: Extreme cases and WLE
« Reply #1 on: November 14, 2016, 01:27:46 AM »
Hi Tong

The default for zero and perfect scores is 0.3 and max-0.3 when using maximum likelihood estimates (this is default but can be changed). WLEs are MLEs correct for the bias at the extreme ends of the scales. You can find more about Warm's correction on the Rasch site: