I now have TBC and was working through the tutorials. In the static processing module, it demonstrated how to separate the baselines into their respective sessions in order to facilitate disabling of dependent baselines. I remember back in the early 90's that this was big topic of debate,i.e., whether to toss the dependent baselines or not. I don't recall any mention of this issue since that time period. I would be interested to hear opinions of this Board. I briefly looked around the internet to see what I could find and quickly came up with two opposing views. One was from the legendary Mike Potterfield. Scroll down to the bottom of the link to the FAQ's. The other was an article by R. Cliff Wilke in "Land Information Systems" vol55,no2,1995,pp99-108.
In my opinion, it probably isn't an issue if you are only using 3 receivers on a routine project. If you are using 5+ receivers on a dam deformation project it probably is an issue. We arrive at the standard "It depends..." response.
What do you think?
My 2-bits is pretty simple:
IF (big IF) you are using a “session” (multi-baseline) processor like PAGES, then it is something REAL. If on the other hand, you are processing EACH baseline separately (like MOST commercial baseline processors do), then I consider ALL baselines to be AT LEAST semi-independent, if not simply independent.
I think that it depends A LOT on just how you configure, OBSERVE, and process/adjust your Network.
Mileage varies of course.
Loyal
> What do you think?
I think weeding out dependent baselines is not worth the trouble. Most of my GPS projects involve multiple (sometimes dozens of) sessions with 4 to 10 receivers. Identifying and eliminating dependent vectors by hand is a major time sink in that scenario. In my opinion, leaving them in doesn't skew the stats anywhere near enough to justify the effort.
If TBC has truly automated the weeding process -- and I'm skeptical that it can do so accurately in a large network -- it might be worth a button-push. But I'm in no rush to try it out. (I'm still trying to figure out why TBC is failing vectors that TGO processed with good statistics.)
Back in the infancy of processing data on an 8088 processor using Ashtech LINECOMP or Trimble TRIMVEC vector processing, and adjusting data using LSA, one of my mentors was Ellis Veach. His philosphy was that declaring one vector from a 3 receiver session dependent was just a guess, as those processors truly processed each possible vector independtly. Best bet: Throw them all in a properly wegithed Least Squares Adjustment and let the adjustment sort out it out. Things like antenna height or wrong point naming issues were quickly identified and resolved.
I always contined to process GPS static vectors indpendently (rather than a group processing session) and adhered to this philosphy as it aided in those blunder detectioins, where a bad antenna height would get "smeared" through a group proccessing session and be difficult to identify, isolate and rectify. Enough reason to stick with that strategy. Thanks Ellis, RIP.....
The problem is that by including the dependent baselines, it skews the statististics in the Chi-square or goodness of fit test. The more receivers you have, the more the statistics are skewed. This results in your error estimates being better than they actually are and thus a true estimate of the errors involved can not be gained. I think the problem most surveyors have with this concept stems from the "more measurements are better" philosophy we generally operate under. A dependent baseline is not another measurement because the software has already used all the "measurements" to compute the independent baselines. A dependent baseline is more or less an inverse between known solutions. The situation is analogous to taking 3 measurements, and using a measurement twice in your statistical analysis. Because of the mathematics, it is in fact much worse than that, but the general idea is the same.
:good:
Ellis Veatch
if there was a beer in my hand, i would toast it to you.
Thanks Ellis, RIP.....
