Kent IS THE POINT
> my point is this, kent. you have a lot of knowledge, but you deliver it in such a down way.
So, you don't really have anything to add to the discussion but you want to object to people discussing the topic? How would you describe that?
Kent does not suffer fools lightly
> It's an argument that doesn't make sense. Either errors are quantifiable (which I believe) or they are not in which case you can't have a tolerance error? How can you measure an unquantifiable subject?
Exactly, Don.
Oh Great!
> So, still no comment on the chi squared test ?
We've been discussing the fundamental problem in your centering test. If the test design is flawed, the statistics are mostly irrelevant.
Oh Great!
At long last I see. You have thrown out the baby with the bathwater, demonstrating your apparent inability to do anything but argue.
I doubt you treat your clients with the same contempt you treat people here. Maybe some afternoon I will have one of my spies take you to lunch and find out how you treat the waiter.
Here is a better question:
given a data set of a random survey job, how would you go about deriving the standard errors for direction, distance, zenith, and all various centering errors from the data itself?
Without being able to "test" the equipment itself?
If you did this, in your humble opinion would it be more or less "true" than the factory spec or some contrived partial measurement at twenty feet?
Can you answer that one cogently and to my satisfaction without being evasive or changing the subject?
This thread is getting funny in that Kent is chiding Perry for not using least squares and meanwhile I am grilling Kent to see if he really knows the inner workings of what he is preaching. And so far, no answers.
Oh Great!
> Here is a better question:
>
> given a data set of a random survey job, how would you go about deriving the standard errors for direction, distance, zenith, and all various centering errors from the data itself?
>
> Without being able to "test" the equipment itself?
>
> If you did this, in your humble opinion would it be more or less "true" than the factory spec or some contrived partial measurement at twenty feet?
The answer, of course is that if you know nothing about the characteristics of the survey equipment, you have no means to derive very good values of the standard errors for directions, distances, zenith angles and instrument and target centering without a highly redundant survey design that is not typical of land surveys.
Even if you had the manufacturer's specification for a particular model of instrument, that would not necessarily be a very good estimator for the instrument used in the survey. The manufacturer's spec is better than nothing, of course.
This is why it's important to independently determine the various standard errors oneself, using a test method that is as efficient as possible. Evaluating centering errors is probably the easiest of the tests, so it's only rational to do it first. It's usually more convenient to have evaluated the centering errors before testing the standard errors of angles and distances. It is unless one has a whole locker full of tripods and tribrachs.
Wow, Kent goes from the rediculous to the sublime.
I can only state that proving plumb of a prism pole and bipod is better than any rod bubble. From decades of data, that is. Error is reduced thereby to the point of adjustment being pointless. A tenth per mile gets the job done for 99% of the boundaries I have done. It simply provides the best value for high level third order surveys without the unjustified expense of equipment simply not required by any minimum standard.
Beware any "surveyor" who has not skill with a plumb line. He is below any mason, carpenter or ditch digger with common practical knowledge and experience.
Oh Great!
So your answer is, you can't, except by the means you have already used, unless you over-design a special survey for the purpose?
How about least squares?
How would you do it with least squares?
Oh Great!
> So your answer is, you can't, except by the means you have already used, unless you over-design a special survey for the purpose?
>
> How about least squares?
>
> How would you do it with least squares?
Someone who wasn't particularly interested in getting realistic standard errors could simply play with the a priori estimates until the standard errors of unit weight for all classes of observations approached unity. This is shown by the uncertainties that would be present in the a posteriori estimates.
Considering that the entire point of having realistic estimates of the standard errors of observations is:
(a) blunder detection and
(b) realistic error propagation,
there is a very large value in as efficiently as possible getting estimates of as many of the various measurement processes with as low uncertainties as can be readily had. This is why testing is a much better strategy than the arbitrary manipulations of weights just to pass the chi square test.
Oh Great!
You admit that your everyday survey design does not gather enough data to derive the standard deviations on a job-by-job basis, that you would have to beef up your field procedure to do so, and that you must resort to contrived measurements not representative of real-world conditions to "estimate" the standard deviations just to get started with your least squares?
Oh Great!
> You admit that your everyday survey design does not gather enough data to derive the standard deviations on a job-by-job basis, that you would have to beef up your field procedure to do so, and that you must resort to contrived measurements not representative of real-world conditions to "estimate" the standard deviations just to get started with your least squares?
Uh, no. Probably most everyday land surveys are not designed to give reliable estimates of the standard errors of the distance and angle observations, let alone centering errors. This is a fact of life and a matter of efficient operations.
This is why in professional practice it is much smarter to get realistic a priori estimates of the standard errors of measurements and observations and to use the adjustment to validate those a priori values. As a matter of professional practice, the goal is to use consistent, predictable measurement processes that can be well characterized by testing. To return to the POB of this thread, this is exactly why a prism pole in a prism pole tripod is the better choice. The process quality can be maintained better and the standard errors are smaller than all other options familiar to me.
Kent does not suffer fools lightly
exactly! I'm pretty sure Kent is the one that was just repeatedly called names by a supposed professional. I think he did a fine job of making his point without returning the schoolyard hijinks.
well, I knew this
was going to be a fun thread, just by looking at the posters. I have to say, I had no idea there was such a die hard crowd of Plumb bob and prism users.
I've actually never used that sort of setup. I didn't see any mention of the random errors that seems would be inevitable as you hold the bob/prism somehow over the point and keep it pointed toward the instrument. I use a geodimeter 650 robot, and it can't track a plumb bob, just that fancy flashing prism. Seems to work just fine for me.
Is it just that bob/prism guys feel like they are doing it "old school" ? Honestly, I'm trying to understand the benefit of this style of surveying. A well adjusted rod with mounted prism and bi/tri pod is a proven repeatable measuring tool.
I know Half Bubble likes stats, maybe he could do the half hour project to check to prove his point?? (Since the original debater clearly thought statistical testing was beneath him.)
Oh Great!
You admit that you are an everyday surveyor, who, for the sake of efficiency, does less than is needed?
In spite of all these disconcerting admissions, you have given me a stellar idea:
My next RSE (redneck science experiment) is going to be with a rod that has no bubble, and plumb it by feel. It works for level rods, right? I wonder how big an issue the center of gravity of the glass will be? I bet it works at least as well as the plumb bob.
TDD will be proud. Don't need no tripod to check no bubble!
Oh Great!
>
> My next RSE (redneck science experiment) is going to be with a rod that has no bubble, and plumb it by feel. It works for level rods, right? I wonder how big an issue the center of gravity of the glass will be? I bet it works at least as well as the plumb bob.
>
usually, everyone rocks a level rod but you knew that.
Keep with the prism/plumb string and just add a good quality decaf coffee. :'(
Oh Great!
> You admit that you are an everyday surveyor, who, for the sake of efficiency, does less than is needed?
Yes, I'm a surveyor every day and I have enough to do that I'm interested in getting high quality as efficiently as possible. Trying to maximize quality while minimizing outlay of time/effort is one of the many keys to professional practice.
Oh Great!
You admit that, having done less than is needed, you still market it as high quality work?
Oh Great!
> You admit that, having done less than is needed, you still market it as high quality work?
No, what in effect I posted was that one of the keys to the practice of the surveying profession is finding efficient ways to obtain high quality results. That's pretty much always been true.
Oh Great!
How do you prove the quality of your results?
Oh Great!
Half,
I hope you look for property corners, they way you are looking to try to catch Kent. So far he has backed every point he made with facts logic and proven survey methods, and I would not try to say he puts out low quality survey work when you admitted that the plumb bob method that you are using starts with 0.02 of error.
How can you defend your survey when another surveyor calls you and says I am not matching your work. Once you tell him you started with know error, well lets just hope the fence is the less expensive thing you have to move.
How many samples of error do you use on a transverse?
Oh Great!
No worries, you are just watching a certain kind of theater. We have to do this every few months and it just happened to be my turn. He has a few things to teach. If we could get him to work on his bedside manner we'd have a winner, send him off as a national speaker ...
I will say that everything I have seen of Kent's has been the finest work. He does some things that very few people do, which is to show his least squares results, discuss how he arrived at his standard errors, talk about repeatable bearings including solars, etc.
I am just giving him a hard time, and if you read the fine print, you will find that often I am using the same recipe as he is just with the equations re-arranged. Like testing your errors vs. deriving them from the chi squared test. You're talking about angels dancing on the head of a pin, either one works. So many people lurk and so few people post that we have to keep these narratives going or a few hundred days from now nobody in the world will remember how to survey.
I spent the day watching the Columbus tutorial videos. Tau criteria, now there's something star*net doesn't have. Time to start a new thread about using least squares to pre-calc a cadaster.
[edit for typo]
