1. I'm surprised that the SEMs in Table 2 are so smal using small sample size.
2. for figure 2, the halo effect was not modeled in Equation 1. It may be still unclear why the Panelist 8 was becoming consistent.
3. I wonder the estimate of slope parameter is so small. Is it significant than .05? The explained variance is small as well.
4. time is an important factor. Panelists may become tired during the process, especially so long test length. Throught the PC power, let panelists use computer to do standard setting. We can record the time spent then. A new IRT model can be used to model this effect.