On 03/20/2014 06:29 PM,
golo...@gmail.com wrote:
> Hi,
>
> I use the same data with the same model in two different versions of
> lavaan. My data consist of ordered factors.
>
> In the 0.5-13, chi squared (DWLS) = 9.428 and lavaan converged normally
> after 18 iterations.
> But in the 0.5-16 version, chi squared (DWLS) = 10.186 and converged
> after 19 iterations.
I would guess this is related to the zero.add= argument which has been
added in 0.5-15. It handles empty cells in bivariate tables. The 'old'
default was to add 0.5 to every empty cell. The new default is to add
0.5 in 2x2 tables only. So the default is zero.add=c(0.5, 0.0): the
first value is for 2x2 tables, the second value is for larger tables. To
get back the old behaviour, use zero.add=c(0.5, 0.5).
But this is just guessing. To be really sure, we need your R script + data.
Yves.