Hi,

sometimes, I'm not sure whether I understand your question in a right way. So, I'm sorry, if I don't answer all your questions completely.

And that such and why 2 and 3?

JAG3D defines four types of points: reference points, stochastic points, datum points and new points. New points are points whose coordinates are unknown. These points will be estimated in the adjustment process. Datums points are new points, too. But in contrast to new points, datum points are defining the geodetic datum of a so-called free network adjustment and they are solving the rank deficit of the normal equation system. Whereas reference points are known points without uncertainties, stochastic points have uncertainties. Therefore, the imperfections of stochastic points are modeled in the stochastic model. During the adjustment process stochastic points get new coordinates, whereas reference points are fixed. Moreover, reference points and stochastic points can be checked by an outlier test.

Type           Correction        Outliertest
New point         Yes                No
Datum point       Yes                No
Stochastic point  Yes               Yes
Reference point   No                Yes

What to do if the network with errors of starting (firm) points is adjustment?

First of all, start the analysis process with a so-called free network adjustment (using datum points instant of reference/stochastic points). In this analysis step, you have to check your observations (leveling, distances, directions,...) and the a-priori stochastic model. If an observation is flagged as outlier, check this observation carefully. In case of doubt, this observation should be disabled. This process is known as Baarda’s data-snooping.
If no more outlier is detected by the stochastic tests, you can move your datum points to a reference point group or to a stochastic point group. In this analysis step, you can check the connection points of your network. If a reference/stochastic point is wrong, JAG3D will highlight an outlier in the point group. The estimated gross error (denotes by nabla) gives an impression of the error of the point. In case of a small error, you can move your reference points to a stochastic point group and regarding this small derivation as a stochastic process. In this way, you have to define the uncertainties of your "firm" points e.g. 1cm. If the error of the point is too large, you have to move this point to a new point group.

It is strange why in New Point influence on defined parameter an expected error of nabla and minimal detectable bias of nabla(a, b) aren't calculated?

Yes, up to now, nabla(a, b) is not calculated for point groups, maybe in a later version.

In Results for New Point, A is confidential interval?

Yes and alpha is used to estimated the interval.

EP, EPavg at you it was calculated as absolyuny value from EP, but it isn't written that isn't absolutely good.

I think, only the absolute value represents a sensible indication. Anyway, a template system is used to generate the report. Feel free to create your own report from JAG3D.

It would be good to place in the field General value for aG.

The critical value is given. Therefore, the test is more traceable, than with alphaG.

In the same place Tprio and Tpost are calculated. As they were calculated?

See equations 16 and 17.

regards
Micha

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