Hi,

1. Calculations manually in Matlab under known formulas to 6 signs after a comma coincide with calculations in version 3.3.20130114, and differ even in 2 sign in the version

I got the same results in v3.3 and v3.4, if I estimate your levelling network, because you are using individual uncertainties. Therefore, the distances are not used to build the stochastic model (click on the picture to enlarge).

To use the distance dependent stochastic model, I remove the individual uncertainties. Here, I can reproduced the differences. I check the changes in the source code. There was a wrong reference, which I fixed now in version v3.4.20130428. Now, the results are equal to each other. Thank you very much! (click on the picture to enlarge)

2. Me interests how to calculate only Kprio, Kpost.

Once again, JAG3D applies Baarda's B-method to estimate these values. The α-value is adapted w.r.t. β and the non-centrality parameter λ.

Using Matlab the critical values is given by the non-central F-distributon e.g. α = 0.1%, β = 80%, m1 = 3, m2 = ∞:

ncfinv(1-0.8, 3, 1000000, (norminv(1-0.001/2)+norminv(0.8))^2)

The function fcdf can be used to get the adapted α.

kind regards
Micha

--
applied-geodesy.org - OpenSource Least-Squares Adjustment Software for Geodetic Sciences