Trouble when using weights, SDs look weird

[yllesaalik]

Hello,
I am using the newest MLwin version probably (2.28) and when trying to use raw weights in two-level modelling, something weird happens.
The standard deviations in brackets behind the coefficients have strange characters, i.e. (1.#IO) or (-1.#IO).
Any idea what it means or how to cope with it?
Without weights the modelling worked fine, coefficients and SDs were there, but then I found out that using weights is relevant. Tried both with ready made models and starting a new one, but still the result is the same- unreadable SDs.
Ülle

[ChrisCharlton]

There is a bug in the calculation of robust fixed part standard errors in MLwiN version 2.28. This will be fixed in version 2.29, which should be available in a few days time. Alternatively they should also work in version 2.27.