I'm continuing to troubleshoot myriad data files gathered for my control network; I reviewed this thread on how to rem out portions of a data line (Thank you Kent):
https://surveyorconnect.com/community/threads/star-net-how-to-rem-out-a-portion-of-a-line.326497/#post-368729
In Post #10 of that thread, Half Bubble commented at length about a method to get Star*net to choke down sometimes errant data from the DC.
He suggested accepting the data as is, but then setting the values in question free (using the asterisk).
I don't have any single values in a line I can test it with, so I'll ask: Is there any difference to an adjustment, between keeping an observation and setting it free, and eliminating it all together? Does a free observation get added to the adjustment, but at some lower weight or something?
Free observations should have no affect on the adjustment.
Jim Frame, post: 386235, member: 10 wrote: Free observations should have no affect on the adjustment.
So is the advantage of keeping the observations, and freeing them, more just for documentation purposes. The alternative (deleting them from the .dat file), might lead to confusion down the road when someone starts comparing the raw DC file with the adjustment files?
I've already run into problems with too much editing of .dat files to correct problems in the raw files. Not sure of the proper way to do this.
I often include unconstrained observations -- mostly coordinates -- in my data files. Sometimes they're part of a group import from which I'll choose constraining values, sometimes they're only for visual comparison to record values, and other times (more often these days) as a means of getting descriptions associated with RTK vectors that come in as g-files, which don't contain descriptions.
I don't recall ever including free angle or distance observations, but I can conceive of situations in which I might want to do so.
We all develop 'rules' for data treatment. Starnet is one place to make absolutely sure the rules are founded in reality.
If things are blowing up (and blunder detection fails to isolate the cause) I run suspect lines free. This leaves the data in to be tested but leaves it out of the adjustment. I use the results to know what to look for in the notes or which specific questions to ask the crew.
I never deviate from my original estimates or remove data without a reason. The field procedures we use provide enough data to eliminate actual bad data without revisiting most of the time. The balance takes time to develop, and changes with every new tool or employee.
As an aside...
The best thing about Starnet us you can learn more about statistics in the real world by using it. If you never lie to yourself it is a great tool. If your pursuit becomes unity and passing your favorite lettered test you're better off without it.
I think a supplied but free value can be used in the initial estimates that get the iteration going. If the algorithm can't get started, you may need to put in a SWAG at some value(s) to help it. That value won't be used to complete the adjustment.
Another use is as mentioned above, is for blunder locating.
You can also cause a distance, bearing, or angle that you didn't measure to be computed for you, sort of a COGO operation piggybacked onto the adjustment.
