I've never done this before. The original levels were done by James C. Conkright, LS3137, in 1984 so I don't have lengths, only turns...
Input file:
[pre]
# Bench Levels by James C. Conkright, LS3137, dated July, 1984
E TBM1 75.85 !
E 57 77.31 *
L 57-771BL 4.62 3
L 771BL-100BL 3.85 1
L 100BL-TP1 -13.31 3
L TP1-TBM1 3.38 2
L 57-SP -1.51 3
L SP-TBM1 0.07 1
L TP1-1421 -2.75 1
L 1421-TBM1 6.14 2
L 771BL-59 9.97 2
L 59-771BL -9.98 1
L 57-PtA -6.26 1
[/pre]
Was the level loop ran trigonometrically.
No, these were differential levels. You could put every turn in but StarNet allows putting in the elevation difference between two points with the number of turns between them.
There is one point in there that he did a trigonemtric level with a Theodolite (I think it is the T1 we have in the equpipment room) but the rest of it I derived from his level notes.
I thought I had some grasp of this subject, but at first glance I'm a puzzled by the error factor 0.995 which is used for the chi-sq test.
How can that be larger than any of the std residuals (0 to 0.9) in the last block of output listing? When the error factor is near 1, I expected the std residuals to be clustered around 1, not all less or much less.
Is it something to do with breaking it into turns?
Duh. I was ignoring the fact that there were only 3 degrees of freedom. Sorry about that.