Mr. Diekman...
very good post..
Lehigh University, Circa 1970's
Statistics was rolled into many courses but was not a separate requirement. That made it harder later on in advanced surveying courses at New Jersey Institute of Technology, circa 2000. 136 credits were required for a BS Engineering, I graduated with 142. At the time Lehigh was one of the few colleges to still require surveying camp for all civil engineers on top of surveying 1. I finally took a 3 credit correspondance course in statistics, as I could not find any advanced 4 credit courses.
Paul in PA
Kevin, 8 Hours Is Not Enough
Larry P's course is more of a review and very good for PDHs.
College level statistics courses are 3 credits even at the community college level. That means 45 hours of class time.
For the technical work of engineers and surveyors that 3 credits is not adequate and many institutions require a 4 credit course covering more advance concepts than most college level statistical work.
Paul in PA
Kevin, 8 Hours Is Not Enough
> College level statistics courses are 3 credits even at the community college level. That means 45 hours of class time.
Really? For a basic statistics course? Wow!
But I know I got really twisted in expectations by my college. Things went by pretty darn fast there. And Caltech's a pretty funky school, and pretty far from what most would consider "normal"... There were statistics courses there, but they would pretty much assume everyone knew what "mean" and "standard deviation" were before the course even started. Next thing you know, you're in partial differential equations... 😛
But like I said, that's a pretty funky school, and they'd get into the esoterica pretty fast. Understanding a simple "standard deviation" is generally enough for most purposes. And that's basically nothing but a measure of how much variability is in your data.
The simple rule of thumb in most cases (where we have what's called a "normal distribution" in the standard cryptic lingo, or a "bell curve" in the lingo most of us learned in middle school) is that about 2/3 of all samples will be within one "standard deviation" of the "mean". I think we all pretty much know what the "mean" is, e.g. what we usually think of as the "average". Then from the mean, 1/3 of all samples will be within one standard deviation lower, and 1/3 will be within one standard deviation higher.
A simple example would be the average height of US males. Using values taken from a Wikipedia article, the mean height of US males is 70 inches, and the standard deviation is 3 inches. This means that if we took the average height of all males in the US, we'd have 70 inches, and approximately 2/3 of all males are between 67 inches and 73 inches. That's really all there is to "standard deviation".
It gets more complicated if you start looking at something other than a "normal distribution", but I've never had such a situation arise in my Surveying work.
> Attorneys should be challenging surveyors not on the math itself, but their use of math over boundary law principles. We should understand these concepts better than an attorney but as a whole I would say we don't.
Reminds me of an ongoing theme in lectures I've heard from Mr. Robillard... 😉
I tend to agree, as well. The math has nice and neat rules. The law does not - it's a chaotic jangling of contradictory concerns. And it's a rare luxury to be on a field crew with a Party Chief who really knows the law, so most of us don't learn that part in the field or through experience.
I agree with what your saying BSA.
I find it interesting that "they" (the royal they) simply took the 2005 wording and substituted the word "Precision" in place of the word "Accuracy".
If I were to venture a guess I would say that by using the word "accuracy" previously there was an implication that a surveyor had to locate a corner to less than 0.07 feet plus 50 PPM. So if I measured to where I think the corner should be and find a pipe or a rock or you name it 0.12 feet away, I needed to set something "closer", thus creating the dreaded "pin cushion"
Most of us would agree that's not correct. 😉
So "they" had to change the word to "precision", thus instructing that the surveyors "measurements" should fall within that tolerance, not the location of a corner.
Did I use enough quotes?
Thanks again for all the input. It is invaluable.
Kevin
Not So Kevin, We Find Many Corners Not Where They Should Be
and it does not affect the accuracy of our survey. Possibly it reflects against the accuracy of the original surveyor.
I surmise the switch from accuracy to precision has more to do with the legal aspects of surveying. As a trained professional I can testify to the precision of the work that leads to my opinions but not to the accuracy. Accuracy may be easily affected by monumentation that I knew nothing about or just could not find with reasonable effort.
Paul in PA
Not So Kevin, We Find Many Corners Not Where They Should Be
Accuracy could be do I have the correct monument or not or it could be the accuracy of my survey within itself. I have 15 redundant observations to the monument at 102.89 feet (+/-0.02') which is reported on the original survey to be at 100.00 feet. My measurement is accurate to 0.02' but the original survey is only accurate to 3' more or less.
I find a lot of monuments in the forest that are slightly leaning downhill. There is a punch mark or cross on top. Since the use of the land is to cut 2' plus diameter trees I don't worry about the couple of tenths difference in the leaning monument; for my purposes I measure to the punch mark in all cases. I could straighten every monument up vertical but maybe it was set that way in the first place besides it will just go back to where it was in short order anyway.
Can anyone tell me if the XL spreadsheet formula for confidence is a fair way to state uncertainty at the 95% confidence level? It's certainly easy to use. The result for redundant coordinate pair observations seems to be about 95% of the spread between min and max as a rule which seems about right.
Properly used, the Excel CONFIDENCE function is correct for one-dimensional gaussian normal (bell-curve) distributions. That would include leveling data and the easting (only) or northing (only) of horizontal measurements. See the help for that function to get more details.
However, it would be misapplied to the magnitude of a horizontal closure. That is always positive and has error interacting in two dimensions. Thus it has a different distribution, resulting in a different factor.
Kevin, you raise good points which generate lots of discussion. That is good.
I don't visit SurveyorConnect every day and my post may be too late - so be it.
Relative Positional Precision is easy if you understand it - duh!
I believe many surveyors have a better handle on relative positional precision than they give themselves credit for in that one needs to be familiar with the capabilities and limitations of the equipment being used. Precision relates to consistent repeat measurements. Consistent results are evidence, but not proof, of good work. Accuracy is more related to coming up with the correct answer as determined by comparision with a standard or higher order work. Accuracy and precision are both important.
Your question relates to precision - that's OK.
But the question I have trouble getting an answer to is "With respect to what?"
Understanding "with respect to what," using calibrated equipment (accuracy), and obtaining consistent results coupled with commonsense will, in many cases, obviate the need for formal application of statistics. But, statistics are also important!
Of course, if measurements do not agree with each other and/or if measurements do not agree with "higher order values" you have a problem that statistics can not fix.
Having said that, statistics does offer some powerful tools of which every surveyor should be aware. But, your are quite correct, learning those tools and concepts is not easy. It takes a lot of work and effort. As a retired educator, I will comment that the rewards are worth the investment in diligent study and developing personal capability.
Jumping off the deep end, I'll suggest the following:
The following paper discusses spatial data accuracy (includes both relative and absolute positioning) and is posted at http://www.globalcogo.com/fsdagsdm.pdf
Not so deep, perhaps is would be best to start with some overheads I've used in seminars and in the classroom over the past 30 years. Judge for yourself. Go to http://www.globalcogo.com/overhds1.html and pull up the following overheads:
PT002
SD001
SD002
SD003
SM001
SM002
SM003
SM004
Of course, there are many other items available as well.
Have fun browsing (you could spend lots of time doing it) and if/when you have any questions, let me know and I'll be happy to help as I can.
Regards always,
Earl F. Burkholder, PS, PE, F.ASCE
Global COGO, Inc.
Las Cruces, NM 88003
www.globalcogo.com
The
Not So Kevin, We Find Many Corners Not Where They Should Be
> Accuracy could be do I have the correct monument or not or it could be the accuracy of my survey within itself. I have 15 redundant observations to the monument at 102.89 feet (+/-0.02') which is reported on the original survey to be at 100.00 feet. My measurement is accurate to 0.02' but the original survey is only accurate to 3' more or less.
>
> I find a lot of monuments in the forest that are slightly leaning downhill. There is a punch mark or cross on top. Since the use of the land is to cut 2' plus diameter trees I don't worry about the couple of tenths difference in the leaning monument; for my purposes I measure to the punch mark in all cases. I could straighten every monument up vertical but maybe it was set that way in the first place besides it will just go back to where it was in short order anyway.
I often apply the idea that accuracy is more akin to "true" or "correct" and precision is a measuring ability.
But as to your example as applied to statistics, if you have a 15 redundant measurements that have a mean of 102.89 (+/-0.02) that is related to a 'precision'. Let's say that the "true" distance at the time you measured it between the monuments you measured between is 103.40. You would not be "accurate" by 0.52' (regardless of the original call and how far off the original monuments were off from the original call).
You or your helper may be setting the glass over a rock, you may have tribrach error that doesn't put you over the point, you have inherent distance error in the machine (regardless of how large or small) you may not have found the original monument that is around 103.4' away from what you measured to (or from)....etc.
Okay; semantics. I see deciding which monument to use is your expertise more than the statistics of your measurements. But the statistics of your measurements is important so that you know how good your reports are.
Just some ramblings, I know most know this.
Tom