chi squared in 2D traverse.

[Claudio Justo]

Hello,

I am using SALSA 1.17.0 (lsasolver) and would like to check whether I am reading the chi-squared output correctly.

For a small network (3 degrees of freedom, APV = 0.251) the .out file shows:

  Chi-squared test lower bound (0.950): 2.605 < 0.251 fail
  Chi-squared test upper bound (0.050): 0.251 < 0.117 fail

The APV (0.251) falls between the two limits (0.117 and 2.605), so the adjustment looks consistent and I would have expected "pass". The example output in the User's Manual (Section 4.2.2) shows "pass" for a healthy fit.

Is the pass/fail label inverted in this version, or am I misreading the test? I also notice two smaller things: the "lower bound" line prints the larger limit (2.605) while the "upper bound" line prints the smaller one (0.117), and the two-sided line shown in the manual does not appear in my output. Attached are the .dat file and the resulting .out  

Thank you for any clarification.

Best regards,
Claudio Justo

[Clark Hughes]

Hi Claudio,

The behavior you're seeing does indeed appear discrepant. We'll take a look as soon as practical to try to understand what's going on and whether this issue may have already been resolved in a soon-to-be-released updated version of SALSA. In the meantime, I'll note there are a couple things about this project that are atypical (and therefore not heavily tested). One is that you're curating the .dat file directly as opposed to building out a .lsa file which the UI can read, display, and of course convert to .dat prior to running lsasolver in the background. Second, you're running a true 2D problem. Even when we play around with 2D problems, we usually work them in 3D but apply vertical constraints; this workflow is numerically inefficient (not an issue for a tiny network) but allows for mapping the network in the UI. I'm not saying you're doing anything wrong; it's just that you're exercising some workflows with very light test coverage. Again, we'll take a look as soon as practical and let you know if we have ideas or resolutions. Thanks for sharing this.

Clark

[Claudio Justo]

Hello Clark,

Thank you for the quick and detailed reply. It is reassuring to hear the behavior looks discrepant to you as well, and I appreciate that you will check whether it is already addressed in the upcoming release.

Your point about the atypical parts of my workflow is well taken, and it also clears up my concern that I might be misusing the software. I plan to try the recommended path: authoring a .lsa project in the UI rather than curating the .dat by hand, and running the problem in 3D with the points constrained in height instead of as a true 2D adjustment.

For context on why my usage is a bit unusual: I am an adjustment computations instructor and I work mostly in Spanish, so I lean on an AI assistant (Claude) to help with the language barrier and with formatting and converting my data into SALSA's input structure. That is partly how I ended up hand-curating the .dat file directly, which I now understand is one of the lightly tested paths.

I will report back if the UI-based .lsa / height-constrained 3D workflow changes what I see in the chi-squared output. Thanks again for looking into this and for the guidance.

Best regards,
Claudio Justo
Facultad de Ingeniería, UNLP