In another thread, it's been stated that if one takes multiple measurements of the same angle or distance, they are to be entered into Starnet as measured (without calculating the mean first). For example, a face left angle of 00-00-01 and face right angle of 00-00-03 would not be entered as a "pre-meaned" 00-00-02.
But in Starnet, you can weight individual measurements, so I'm curious if you can enter such "pre-meaned" measurements, apply a greater weight to that line, and get the same results. I've tried a few combinations, but a clear correlation is not emerging. Would the weight be the ratio between a standard mean and a least squared mean?
Or would the weight factor be the std. error (say 10" in this case), divided by the number of meaned measurements?
> In another thread, it's been stated that if one takes multiple measurements of the same angle or distance, they are to be entered into Starnet as measured (without calculating the mean first). For example, a face left angle of 00-00-01 and face right angle of 00-00-03 would not be entered as a "pre-meaned" 00-00-02.
Actually, the mean of directions taken Face Left and Face Right (the latter being reduced by 180-00-00) is only ONE observation/measurement. So, enter the mean of the two faces as one observation. Do not treat the Face Left and Face Right directions as separate observations. Your total station only has one circle reading point, so the directions on each face contain circle eccentricity errors that are canceled in the means of both faces.
:good:
I agree with Kent.
To explain what I think is the thrust of your question, the variance of the meaned value reduces by the number of sets, so the std deviation reduces by the square root of the number.
If you do a D+R set and then another D+R set under similar conditions (best with re-centering if that is significant to the error), the proper std error to give Star*Net for the mean of the two sets is 1/sqrt(2) = 0.7 times the std error of one set.
> > In another thread, it's been stated that if one takes multiple measurements of the same angle or distance, they are to be entered into Starnet as measured (without calculating the mean first). For example, a face left angle of 00-00-01 and face right angle of 00-00-03 would not be entered as a "pre-meaned" 00-00-02.
>
> Actually, the mean of directions taken Face Left and Face Right (the latter being reduced by 180-00-00) is only ONE observation/measurement. So, enter the mean of the two faces as one observation
OK. I picked the wrong example. To further my understanding of what Starnet is doing under the hood, I'm still curious about MULTIPLE measurements. Let's say they're BOTH Face Left. Is meaning them and entering them once the same as entering them each?
> OK. I picked the wrong example. To further my understanding of what Starnet is doing under the hood, I'm still curious about MULTIPLE measurements. Let's say they're BOTH Face Left. Is meaning them and entering them once the same as entering them each?
With your instrument, don't use directions that aren't taken as the mean of both faces.
Least squares survey adjustments only work when the random errors in the observations and measurements are well characterized. That would not be true of directions on only one face. They will contain various errors that would have been canceled in the mean of Face Left and Face Right.
If I turn say, 4 sets (1 set=FL+FR) of angles between two points (or 4 sets of directions between multiple points), I use the mean of the 4 sets and the computed standard deviation rather 4 different angles (mean of D+R) in the adjustment. However, if I do two separate sets of sets, then I do enter them as separate observations. I do not use Star*Net but rather Geolab, but I imagine the functionality is the same.
> If I turn say, 4 sets (1 set=FL+FR) of angles between two points (or 4 sets of directions between multiple points), I use the mean of the 4 sets and the computed standard deviation rather 4 different angles (mean of D+R) in the adjustment. However, if I do two separate sets of sets, then I do enter them as separate observations. I do not use Star*Net but rather Geolab, but I imagine the functionality is the same.
If the multiple sets are from the same setup and the targets are at close range, yes, it would be more rigorous to form the means of all sets as observations and just decrease the standard error of a direction accordingly.
In this case, I don't think we're quite there yet.
> With your instrument, don't use directions that aren't taken as the mean of both faces.
For control surveying, that is. Single face directions (with the vertical circle properly indexed) are typically fine for mapping coarse features like trees and other topo shots.
> > With your instrument, don't use directions that aren't taken as the mean of both faces.
>
> For control surveying, that is. Single face directions (with the vertical circle properly indexed) are typically fine for mapping coarse features like trees and other topo shots.
I'm definitely interested in learning control surveying at this time. What John said:
"If you do a D+R set and then another D+R set under similar conditions (best with re-centering if that is significant to the error), the proper std error to give Star*Net for the mean of the two sets is 1/sqrt(2) = 0.7 times the std error of one set."
is precisely what I calculated, based on a comment Kent made in another lengthy thread.
I started this line of question, partly to get around the limitations of the demo version of Starnet, which I'll bust on my upcoming "Kent II" test (4 orthogonal close up targets and five further away). The thinking was I could "pre-mean" sets of measurements prior to entering into Starnet.
But that led me to try to understand exactly what Starnet is doing, and how it's different from what Ghilani/Wolf describe, which is a pretty standard calculation of standard deviation.
> I started this line of question, partly to get around the limitations of the demo version of Starnet, which I'll bust on my upcoming "Kent II" test (4 orthogonal close up targets and five further away). The thinking was I could "pre-mean" sets of measurements prior to entering into Starnet.
You always have the option of computing the standard error of centering of the total station separately, but it would be simpler just to follow the program that I outlined earlier and automatically correct the directions for the instrument center for each setup while solving the standard error of a direction taken as the mean of face left and face right.
If you want to solve the standard error of centering separately, you just measure sets of directions to the four inner targets, taking the means of Face Left and Face Right as the observations. You'll need to set up the instrument at least ten times, rotating the tribrach about 60 degrees between setups, and solve the coordinates of the instrument station in smaller batches, but it is definitely doable.
With the standard error of instrument centering known, you can then adjust the sets of directions to the more distant targets, rotating the instrument circle between sets as was previously discussed, but accounting for the centering errors by the value of the standard error of instrument centering, rather than explicitly solving the centering error values as would be done by the program I outlined.