Thread on meaning multiple sets of observations

yes, use each individual measurement. it is a testament to precision

Actually, if you take, multiple sets, I'd use the mean of all of the sets if they are all taken from the same setup to the same targets, i.e. centering errors of the instrument and the targets are identical.

You would have to change the standard error of the mean of three sets to be s/SQRT(3), where s = standard error of an angle taken as the mean of Face Lt and Face Rt.

If your least squares adjustment software doesn't allow you to vary the standard error of an angle, then you're SOL, but if you're using Star*Net, you have two options, one of which is two just explicitly state the standard error on the data entry. Another way is to create an instrument in the instrument library that is, say "TSx3" that has a standard error equal to the mean of three angles.

.INSTRUMENT TSx3

as a line in the input file instructs Star*Net to consider all data lines that follow to have the standard errors specified for TSx3.

(Wow, my keyboard certainly seems to have more than the usual problems after I drove to and from Houston today.) Make the above:

Actually, if you take multiple sets, it would be better to use the mean of all of the sets if they are all taken from the same setup to the same targets, i.e. centering errors of the instrument and the targets are identical, assuming the distances from the instrument to the target are short enough that the effect of centering errors is significant.

You would have to change the standard error of the mean of n sets to be s/SQRT(n), where s = standard error of an angle taken as the mean of Face Lt and Face Rt.

For example, if the standard error of an angle taken as the mean of a set of directions observed Face Lt and Face Rt is 3.0", then the standard error of the angle taken as the mean of two such measures would be 3.0"/SQRT(2)= 2.1".

If your least squares adjustment software doesn't allow you to vary the standard error of an individual angle, then I think you're SOL unless you use the exact same procedure to measure every angle in the network subject to adjustment or error analysis. If you're using Star*Net, you have two options, one of which is to just explicitly give the standard error on the data entry. Another way is to create an instrument in the instrument library named, say, "Elta50x3" that has a standard error equal to the mean of three angles.

The following command line in the input file :

.INSTRUMENT Elta50x3

instructs Star*Net to consider all data lines that follow to have the standard errors specified for what you designated as "Elta50x3" in the Instrument Library.

Interesting stuff.

> Actually, if you take multiple sets, it would be better to use the mean of all of the sets if they are all taken from the same setup to the same targets, i.e. centering errors of the instrument and the targets are identical, assuming the distances from the instrument to the target are short enough that the effect of centering errors is significant.
>
> You would have to change the standard error of the mean of n sets to be s/SQRT(n), where s = standard error of an angle taken as the mean of Face Lt and Face Rt.
>
If I understand correctly, as long as I lower the standard error appropriately (in my case it will be around 2.25") associated with the combined 3X BD/FD-FR/BR observations, I can use the result in Star-net or other LSA program.

This might actually work for me, as Jim Frame reports in my other closely related thread on the subject, that SurvCE MAY be doing the meaning and reporting the results as a SS record. I'm still trying to figure out what TraversePC is doing with the data.

Before I'm done, I'm sure I'll be putting these observations into both programs and comparing the results. If nothing else, in the meantime, I know I can rely on Starnet to make the adjustments on these control points, and use it's resulting coordinates going forward.

Interesting stuff.

This is your meaned angles:
[pre]
--SS,OP102,FP101,AR123.3646,ZE91.0747,SD239.444167,--

--SS,OP102,FP301,AR262.2845,ZE83.3347,SD355.145500,--hill
[/pre]

The -- comments the line out. SurvNet would use all the observations, I think. In StarNet you can either enter all of the observations or use the mean. If you are using a converter it is pretty easy to tell which it is doing. A meaned record will only have one DV and one M line but if all the measurements are used then there will be several DV and M lines.

Edit-well heck, the meaned SS lines are both short 10" from my calc of the mean of the 6 angles. The reduced angle right would be the same - 138°51'59". I'm not sure why it's doing that.

Interesting stuff.

> This is your meaned angles:
> [pre]
> --SS,OP102,FP101,AR123.3646,ZE91.0747,SD239.444167,--
>
> --SS,OP102,FP301,AR262.2845,ZE83.3347,SD355.145500,--hill
> [/pre]
>
> The -- comments the line out. SurvNet would use all the observations, I think. In StarNet you can either enter all of the observations or use the mean. If you are using a converter it is pretty easy to tell which it is doing. A meaned record will only have one DV and one M line but if all the measurements are used then there will be several DV and M lines.
>
> Edit-well heck, the meaned SS lines are both short 10" from my calc of the mean of the 6 angles. The reduced angle right would be the same - 138°51'59". I'm not sure why it's doing that.

That may be because I discovered (OK...stumbled on:-) ) a bug in the program. Here's what they said:

>TPC was correctly computing the mean of the individual sets but was retaining the F1 record of each set and meaning those. So if you were to manually compute a mean for each angle set, then average them, you would get a different mean than what TPC gave you. We understand why some folks want the [x] Keep F1 option and other don't.

I can honestly say I have no clue what that means (is that a pun?).

Interesting stuff.

I don't understand that either, what is the "Keep F1" option?

Interesting stuff.

> I don't understand that either, what is the "Keep F1" option?

Here's what they say:
>The " [x] Keep F1" option is a user preference that affects what data TPC takes away from each angle set. It can use the mean from the set, or just use the F1 record from the set. It's in there because people have asked for it over the past 28 years.

I understand this statement but it begs the question: if you're going to use the F1 observation only, why take sets at all?

They go on to say:

>Our traverse view displays inverses based on the coordinates or geodetic positions unless you tell it to display raw data. In your traverse, you compute point 301 with BS=101, OP=102, FS=301. Then you compute point 101 with BS=102, OP=301, FS=101. So point 101 now gets recomputed, which changes the inversed horizontal angle displayed for 301. In TPC, a FS or SS computes a new position for that point. So you could have used a new point label like 101:1 and put =101 in it's point description. Now you could easily check the inverse between 101 and 101:1 in the field or in the office and the =101 would have told TPC to create a redundant observation for 101 in the LS analysis and point 101 would not have been recomputed and the inversed horizontal angle at 301 would match the raw angle.

I think they're trying to make the point that multiple sets measuring one angle between two points isn't as good as using another point in the network from which to measure to each of the others...in other words, a network rather than a single line traverse.

I think in my case, since I'm going to do that anyway, and am attempting to learn practices that result in the highest quality work, time consuming as it is, I'll continue to take three sets, use the mean from them, apply the lower standard error Kent suggests in LSA, and use the resulting adjustments from that.

Here is the thread you're referring too. I have read it a couple of times and have always found it on google.

https://surveyorconnect.com/index.php?mode=thread&id=85236#p85268

for some reason the link brings you to about 5 or 6 posts down instead of starting at the top.