23Likes
01062017, 07:32 AM

Beartooth Regular


Join Date: Mar 2015
Location: Colorado
Posts: 54


OCW Testing
I have begun an OCW Test for my 7mm STW
I wanted to get opinions on using Ontarget or other programs for measuring accuracy?
With ontarget should i take the measure of all 5 shots or the closest three?
Should I go back and shoot 3, 5, or 10 of the best powder charges?
I have attached the Ontarget from one charge to see what you all think.
I appreciate any and all advice

01092017, 11:10 AM


Beartooth Regular


Join Date: Apr 2006
Location: Arizona, USA
Posts: 2,271


I have obtained my best results, when using the OCW method, by following Dan Newberry's instructions regarding how to shoot the series. OnTarget can be helpful in nailing down the center or 'average' point of impact for each group, but remember that the initial OCW workup does not contemplate measuring or paying attention to vertical or horizontal spreads, apart from helping an experienced eye identify what Dan calls a 'scatter node.'
I have read a few times a thing attributed to Walt Berger: 3round groups tell you more about the load, whereas 5round groups tell you more about the shooter. Don't extend that too far, nor try to overanalyze it. Just know that the point is that most shooters are the main contributors to their own poor results. ;)
I would not advise discarding any shots in a group. That is how one 'cherrypicks' data to support their desired conclusions, rather than letting the data direct them to the right conclusion. The only exception _may_ be if you called a bad shot _before_ you saw it on target, and correctly called the direction and approximate magnitude of your 'miss.' A simple 'oh, I don't think that was a very good shot,' without more specificity, IMO may drive one to reshoot the entire sequence, or just to practice more in general, but to me it is not sufficient to start discarding individual shots. ...especially if it happens in a lot of one's load workups... ;)
__________________
"A government big enough to give you everything you want is a government big enough to take from you everything you have." Gerald Ford

01112017, 02:32 PM


Super Moderator


Join Date: Jan 2005
Location: Hilliard, Ohio
Posts: 12,151


Good reply. For further reading, understand the Texas Sharpshooter Fallacy. I fear Walt Berger's comment suggests he may have fallen at least partly victim to this. Many folks don't realize that the larger the group shot count is, the larger the group will tend to be even from the same rifle with the same ammunition fired the same way each time, as from a machine rest. This is because you give more opportunities for less probable (further out on the bell curve) shots to occur, but that doesn't mean they aren't legitimately part of your bell curve distribution. Just that you are allowing more of the tardy and black sheep and alternative lifestyle relatives into the party. Larger groups have the advantage that the measured mean gets closer to what an infinite number of shots would produce.
__________________
Nick
__________________________
Orange Hat Family Member
CMP Certified GSM Master Instructor
NRA Certified Instructor
NRA Benefactor Member
"First contemplation of the problems of Interior Ballistics gives the impression that they should yield rather easily to relatively simple methods of analysis. Further study shows the subject to be of almost unbelievable complexity." Homer Powley
Last edited by unclenick; 01112017 at 02:35 PM.

02082017, 03:55 AM

Registered User


Join Date: Jan 2016
Location: Northeast OHio
Posts: 12


"I wanted to get opinions on using Ontarget or other programs for measuring accuracy?
With ontarget should i take the measure of all 5 shots or the closest three? "
,t
When you get below 1MOA it is difficult to tell a .1MOA improvement. OnTarget helps. I look at my targets at the range then go home and enter them into On Target. Entering them into OnTarget forces me to spend more time with the target. This is probably better spent time that playing with Quickload or Reading for Thought.
I have used 3, 4,5, and larger groups. I can't say that one number is better than the next. The attached target shows 4 shot groups. Why 4 shot groups? It was a convenient number. I have a group of 50 cases I reload at one time. I have 12 bullseyes on the target. 12 goes into 50 four times with two left over for fouling shots.

02082017, 09:41 AM

Beartooth Regular


Join Date: Jan 2017
Location: Oakland Co MI
Posts: 259


Quote:
Originally Posted by brantfred
"I wanted to get opinions on using Ontarget or other programs for measuring accuracy?
With ontarget should i take the measure of all 5 shots or the closest three? "
,t
When you get below 1MOA it is difficult to tell a .1MOA improvement. OnTarget helps. I look at my targets at the range then go home and enter them into On Target. Entering them into OnTarget forces me to spend more time with the target. This is probably better spent time that playing with Quickload or Reading for Thought.
I have used 3, 4,5, and larger groups. I can't say that one number is better than the next. The attached target shows 4 shot groups. Why 4 shot groups? It was a convenient number. I have a group of 50 cases I reload at one time. I have 12 bullseyes on the target. 12 goes into 50 four times with two left over for fouling shots.

Regarding your photo.
I too put stickers on my targets. I use the ones by birchwood casey that have the alignment lines. Given that someone else uses stickers too... I figured it was time I cooked up some nonsticker targets that did almost the same thing.
So here you have it. The PDF has a page with a 1/4 grid and without the grid.. I based it off the target you used from reloaderslog. I'll probably make another in Tabloid (11"x17") with neon green instead of yellow. Bulls will be slightly bigger for my less accurate hunting rifle load development.
Here is an small view of what it looks like:
I have updated the PDF with a few alignment tweaks that I noticed after first publishing
Last edited by Dimner; 02082017 at 11:48 AM.
Reason: Updated PDF

02082017, 10:37 AM

Registered User


Join Date: Jan 2016
Location: Northeast OHio
Posts: 12


'I too put stickers on my targets. I use the ones by birchwood casey that have the alignment lines. Given that someone else uses stickers too... I figured it was time I cooked up some nonsticker targets that did almost the same thing."
Great Minds think alike! You've taken it a step further, and I've downloaded your pdf. The grid which I can see through my scope at 100yards is right on each bull'seye. This will make optics adjustments easier. Thank you for sharing.

02082017, 10:54 AM


Beartooth Regular


Join Date: Dec 2008
Location: Alamogordo, NM
Posts: 7,006


Quote:
Originally Posted by MZ5
I have obtained my best results, when using the OCW method, by following Dan Newberry's instructions regarding how to shoot the series. OnTarget can be helpful in nailing down the center or 'average' point of impact for each group, but remember that the initial OCW workup does not contemplate measuring or paying attention to vertical or horizontal spreads, apart from helping an experienced eye identify what Dan calls a 'scatter node.'
I have read a few times a thing attributed to Walt Berger: 3round groups tell you more about the load, whereas 5round groups tell you more about the shooter. Don't extend that too far, nor try to overanalyze it. Just know that the point is that most shooters are the main contributors to their own poor results. ;)
I would not advise discarding any shots in a group. That is how one 'cherrypicks' data to support their desired conclusions, rather than letting the data direct them to the right conclusion. The only exception _may_ be if you called a bad shot _before_ you saw it on target, and correctly called the direction and approximate magnitude of your 'miss.' A simple 'oh, I don't think that was a very good shot,' without more specificity, IMO may drive one to reshoot the entire sequence, or just to practice more in general, but to me it is not sufficient to start discarding individual shots. ...especially if it happens in a lot of one's load workups... ;)

Yes very good. I learned more than just a couple things reading that.
__________________
Stretch ®
 "A people that elect corrupt politicians, imposters, thieves, and traitors are not victims, but accomplices"  George Orwell
 "In an age of universal deceit, telling the truth is a revolutionary act"  George Orwell
 "If you contribute something, you don;t contribute enough. If you contribute nothing, you receive everything."  Andrew Wilkow
DON'T BE SCARED© †

02092017, 12:47 PM


Super Moderator


Join Date: Jan 2005
Location: Hilliard, Ohio
Posts: 12,151


Dimner,
Read Erick Steckers article on bullet jump. Not to be discouraging, but I suspect you are wasting your time with such small incremental changes to seating depth. I wouldn't go as far as to treat your bullets like VLD's, but I would suggest for secant ogive bullets you try steps about ten times larger and for tangent ogives about 7 times larger.
The difference between 3,4,5 and 10 shot groups is where the confidence limits lie. If you shoot only 3shot groups and your sources of error are all random (for simplicity here, we'll say they are), where the average group size is 100%, then the expectation is that, on average, 19 out of 20 groups will fall inside the following limits, and the 20th will fall outside either the upper (be larger) or lower (be smaller) limit:
3 shots: 40.6%247% of average size
4 shots: 56.4%177% of average size
5 shots: 65.2%153% of average size
9 shots: 79.4%126% of average size
You can see from those numbers that your certainty gets much better as you go to a larger sample size. The reason for this is, as I mentioned before, a larger group size gives more opportunities for less likely shots that are further from the average location to occur, so groups get bigger, but by including a certain number of less probable rounds, the number that are even less probable than that diminishes exponentially, so you get closer to the limit of how badly the gun can group.
Here is a table showing how much average group size will grow with shot count. You can see that if, like an ammunition manufacturer commonly does, you fire a 100 or a 200 round sample for accuracy testing, the group will average about 2.72.8 times bigger than the average 3shot group. That is just due to random distributions tending to put more close to the center so you are less likely to get a small group to show the possible scatter range. You can see the average 5shot group will be about 1.3 times bigger than the average 3shot group, which likely fooled Walt Berger. He's right in the sense that the bigger the group, the more chance there is for a significant shooter contribution, but what I show below is just random and not shooter dependent.
This shows the range of error for 95% confidence (19 out of 20; a more exact number is 21 out of 22, but you don't easily resolve that error in these group sizes, so it is the statistical equivalent of picking nits).
In my example of error for individual sizes of groups, you saw how much tighter 9 shots is than 3. For this reason, I evaluate OCW targets by running averages of 9 shots to spot clumping. The following graph shows real data from a fellow on another board. The purple diamonds are the mean centers of his 3shot groups, but the yellow round dots show the running average of sequential 9shot POI's from the same data. You can see how much easier it is to spot where the groups POI's tend to clump together than it is when you just look at the purple diamonds. The straighter line made by the 9shot averages reflects the greater reliability of larger numbers of shots in a group to tell you what is actually happening.
Here is a simulation. It is a group of 1000 shots randomly distributed as a bivariate normal distribution (typical for zero wind machine rest gun). The first shots 1 through 3 of the 100, the next 3, 4 through 6, and the third 3, 7 through 9 are shown. The yellow cross is the average for the whole 1000 shots, while the green X's are the average for each three shot group. So the question is, if you shoot just three, how much is it telling you about the next 1000 rounds through your barrel and where the average POI will be? How much do your 4 shot groups resemble this kind of random variation and how much have they really told you about whether or not you've found your load sweet spot?
If I redo that for 4 shots or 5 shots, you will see the average group size grow, as the first image in this post shows, but the average location will get closer to the 1000 shot average location, giving you a better idea what the next 1000 shots may do. Note that even the first target of 3, though the biggest group, has no shot nearly as far out as a couple of the shots in the 1000 shot total group includes. Yet, if we divide the whole 1000 into sets of three, each of the most extreme shots in the bigger group will turn out to be part of one of those threes (unless the very last remainder shot was one of them; but what are the odds of that?).
__________________
Nick
__________________________
Orange Hat Family Member
CMP Certified GSM Master Instructor
NRA Certified Instructor
NRA Benefactor Member
"First contemplation of the problems of Interior Ballistics gives the impression that they should yield rather easily to relatively simple methods of analysis. Further study shows the subject to be of almost unbelievable complexity." Homer Powley

02092017, 01:15 PM

Beartooth Regular


Join Date: Jan 2017
Location: Oakland Co MI
Posts: 259


unclenick.
That post is truly fascinating. I did read through the texas sharp shooter falicy. It was a very interesting read and makes alot of sense. I think there was a paragraph about that in one of my text books back in college, but not nearly in as much detail or with as many helpful examples.
Oh, and I actually wasn't the one doing the OCW testing or testing bullet seating... I just liked the guys target.
but......
you have a knack for timely advice. I'm very close to doing my own OCW testing and playing around with seating depths.
Please keep checking back to this thread, because I already have questions for you based on your post. I have read it 3 times, and each time I find an answer in the text, but it brings up another.
I have to get on the road home from work, but I'll have some questions tonight for you.

02092017, 02:01 PM


Super Moderator


Join Date: Jan 2005
Location: Hilliard, Ohio
Posts: 12,151


Glad it helps. I have Excel files that do the evaluation. I think I'm going to go ahead and buy the Ontarget author's TDS software, too, though I don't know if it will do exactly what I do to evaluate a target. Shooting stats are about trying to see through the noise, which isn't always easy.
If you need input and I don't see the post in a timely fashion, just PM me a link back to it so I know to take a look.
__________________
Nick
__________________________
Orange Hat Family Member
CMP Certified GSM Master Instructor
NRA Certified Instructor
NRA Benefactor Member
"First contemplation of the problems of Interior Ballistics gives the impression that they should yield rather easily to relatively simple methods of analysis. Further study shows the subject to be of almost unbelievable complexity." Homer Powley

02102017, 11:13 AM

Registered User


Join Date: Jan 2016
Location: Northeast OHio
Posts: 12


"Read Erick Steckers article on bullet jump. Not to be discouraging, but I suspect you are wasting your time with such small incremental changes to seating depth. I wouldn't go as far as to treat your bullets like VLD's, but I would suggest for secant ogive bullets you try steps about ten times larger and for tangent ogives about 7 times larger."
I think it was my target with the .003 changes in jump that you were referring to. Your post is thought provoking. Many questions arise in my mind. The first is, Am I reading noise when I see that different groupings at different seating depths? The second is, If I move the distance between seating depths from .003 to .020? will I get a different picture of things? Should I move to .030 distance between seating depths?( I currently shooting Sierra 2155's) The third is, Would an upgrade to the OnTarget TDS where you can take multiple groups and make a plot of the group size and the distance between the groups be worth while.?
I shoot at a local ODNR range which opens on March 1st and I received a rifle back from the gunsmith in January. The first session will be to work up from a starting load, but the second session isn't defined yet. When I get to seating depths, I will give larger distances between seating depths a try.
Last edited by brantfred; 02102017 at 11:16 AM.

02102017, 11:21 AM

Beartooth Regular


Join Date: Jan 2017
Location: Oakland Co MI
Posts: 259


Okay ... I have some questions. Hope they make sense.
Quote:
Originally Posted by unclenick
Read Erick Steckers article on bullet jump.

Yes, I did read that a while back when I first learned about OCW. I realize it's information about a specific VLD bullet. What's the consensus about flat based bullets? I guess the one's I'm using are neither secant or tangent ogive. Just regular old spitzer, or have not been named secant or ogive by the manufacturer. Case in point. I'm about to do a seating depth test on sierra's .224 #1310 ProHunter SP. I was going to use the Stecker article's method, but if there is a better way for the #1310 I'm all ears.
Quote:
Originally Posted by unclenick
The difference between 3,4,5 and 10 shot groups is where the confidence limits lie. If you shoot only 3shot groups and your sources of error are all random (for simplicity here, we'll say they are), where the average group size is 100%, then the expectation is that, on average, 19 out of 20 groups will fall inside the following limits, and the 20th will fall outside either the upper (be larger) or lower (be smaller) limit:
3 shots: 40.6%247% of average size
4 shots: 56.4%177% of average size
5 shots: 65.2%153% of average size
9 shots: 79.4%126% of average size

Totally makes sense, as does the graph showing shot size increasing with the # of shots.
Based on what you have put together a 9 shot group looks like the right way to go. but... (yeah you knew it was coming) what about the very budget conscious load developer? Given, we are talking statistics and probability here (or at least I think we are). I know the numbers are the numbers, there is no way to change that. A 9 shot group is going to give a much closer result to the true average than any number below it.
But when you add up the components needed for load development from start to finish you are looking at alot of bullets, powder, primers, and case life. For purposes like mine, I need to consider the cost of the entire load development process as well.
Is there or can there be created a method for soup to nuts load development that balances cost with confidence?
Many will say budget and savings and reloading are not in the same dictionary, but for me they are. Sure if I am competing I'm not worried about the number of bullets I use until I get my load right, but Mr. Dimner is not a competitor and is on a fixed budget
I used the term method above because it's not all about the number of shots in a group or even the OCW process on its own. OCW is a great process that saves on components to find a node that is stable, but to use that process, one has to have a starting point. In my method, I make loads to meet a velocity range I am trying to achieve or a terminal energy estimate at the max range the load will be used. Still have not gotten quickload to estimate better than 811% for velocity. So i have to use actual velocity measurements. Then I use that range to setup my 6 group OCW test. Sometimes, I do the OCW test twice just to be sure. Then I find the accuracy node and scatter node, then move to working on bullet seating length. Add 9 shot groups to all of these steps and it's out of my budget. Can this method be optimized either with an educated reduction in shot number and/or a change in the overall method?
This is a total rabbit hole I realize. Just things I think about when developing loads. We can skip this portion of the discussion if you prefer.
I also have a question about the group center location graph, but I don't think I will be able to get to it until tonight.

02102017, 11:50 AM

Beartooth Regular


Join Date: Mar 2013
Posts: 1,525


Quote:
Originally Posted by Dimner
Totally makes sense, as does the graph showing shot size increasing with the # of shots.

It is intersting to note that the true mean 'dispersion' =
sqrt(n/n1)*(measured dispersion)
Example:
Measured = 1 inch, 3 shots
True = sqrt(3/2)*(1 inch) = 1.22 inch
Measured = 1 inch, 9 shots
True = sqrt(9/8)*(1 inch) = 1.06 inch
__________________
"The purpose of computing is insight, not numbers." Richard Hamming

02202017, 06:36 PM


Super Moderator


Join Date: Jan 2005
Location: Hilliard, Ohio
Posts: 12,151


F2G1D,
My past assumption was always that √n/(n1) was an attempt to compensate for the underestimate that a small sample tends to provide, but Denton Bramwell, who makes his living with statistics, said that's reading too much into it. He said it is actually an artifact of Fisher's ANOVA that worked out to make the math neat. This article on unbiased estimation of standard deviation is apropos.
Brantfred,
You are right. It was your target I was referring to.
I've repeated this in a number of places before: The late Dan Hackett, a benchrest competitor, wrote in the 1995 Precision Shooting Reloading Guide about a 40X he had in 220 Swift that he could not get to shoot well (by his standards). He said the best 5shot group he had from it was about ⅜" and it normally shot more like ½" and sometime more like ¾" (note how closely this fits the range in my table for 5shot samples). At the time he loaded for all his rifles with 0.020" jump because everyone "knew" that was best, same as they'd known it was best to use 0.025" and 0.030" a few years before, and the way they'd known that touching the lands was best a couple of decades before that. Anyway, one day he loaded some rounds with a bullet (don't recall him saying which) but then switched to a Nosler bullet that was shorter. So he adjusted his micrometer seating die to compensate for the length difference, but accidentally turned it the wrong way, so instead of the jump being 0.020" off the lands, it was now going to chamber 0.050" off the lands. He had 20 rounds loaded before he noticed the error. He debated pulling the bullets, but decided just to shoot them in practice, but not expect much. To his surprise, those 20 rounds gave him two 5shot groups of ¼" and two true bugholes in the 1's.
If you look at Dr. Lloyd Brownells 1965 study of pressure, he noted its change with seating depth could be substantial. He attributes it to gas escaping between the time the bullet starts to move and the time the bullet obturates the bore. In general, the bigger the jump, the larger the annular gap between the bullet and throat is when the gas begins to escape around it, so the greater the volume that gets out before obturation and the longer it takes for peak pressure to return to its climb toward the peak, and the faster the bullet completes entering the bore and continuing expansion of the powder containment space. Brownells jumps are large because his bullet is round nosed, which form includes an ogive with a very gradual taper, so it has to move further to change the size of that annular gap much.
The group overlaying feature of the TDS is one of my reasons for interest in it. IRRC, it has a 15 day free trial, so you can download without buying and try it out. I believe the free trial includes full feature operation. I downloaded it but have been waiting to install it until I can run a full test load cycle with it within that trial period.
Dimner,
If you look at the third plot in post #8, the locations of each purple diamond is where the center of a 3shot group was compared to the point of aim. There are 8 of them, so this was a 24 shot test. He only needed the first six (18 shots total) as it turned out, but he had no way to know that at first.
The nine shots I used in that evaluation were not separately fired groups. Rather, the first yellow dot is the average POI for the first three sets of 3shot groups. These were the three 39.8 grain loads, three 40.1 grain loads, and three 40.4 grain loads, all combined to find the average center of all nine of those shots. The next yellow dot drops the first group of three loads (39.8 grains) and picks up the fourth, and then averages the resulting nine POI’s, so it is the average location for the 40.1, 40.4, and 40.7 grain loads. The next yellow dot drops the 40.1 grain POI’s and adds the 41.0 grain POIs and gives the total average for 40.4, 40.7, and 41.0, and so on. That’s how a running average works; dropping the earliest and adding the next and keep all that were between them. It makes the sweet spot easier to identify because combining nine results in something closer to the average POI an infinitely large group would produce.
If you have Excel, the Data Analysis Pak that comes with it has a version of Student’s Ttest in it. This is a test of how much confidence you can have that two different size groups are genuinely different as compared to just randomly different. It starts to get believable at confidence levels of 90% or 95%, as are commonly used in industry.
__________________
Nick
__________________________
Orange Hat Family Member
CMP Certified GSM Master Instructor
NRA Certified Instructor
NRA Benefactor Member
"First contemplation of the problems of Interior Ballistics gives the impression that they should yield rather easily to relatively simple methods of analysis. Further study shows the subject to be of almost unbelievable complexity." Homer Powley

02202017, 07:25 PM

Beartooth Regular


Join Date: Mar 2013
Posts: 1,525


Quote:
Originally Posted by unclenick
He said it is actually an artifact of Fisher's ANOVA that worked out to make the math neat.

This analysis came later  Exterior Ballistics, Hermann, 1935.
Here, Measures of Dispersion
you can see that S = SN* sqrt[n/(n1)]
__________________
"The purpose of computing is insight, not numbers." Richard Hamming
Last edited by F2G1D; 02202017 at 07:49 PM.

02212017, 10:51 AM

Beartooth Regular


Join Date: Mar 2013
Posts: 1,525


__________________
"The purpose of computing is insight, not numbers." Richard Hamming

02212017, 01:16 PM

Beartooth Regular


Join Date: Mar 2013
Posts: 1,525


__________________
"The purpose of computing is insight, not numbers." Richard Hamming

02212017, 05:50 PM


Super Moderator


Join Date: Jan 2005
Location: Hilliard, Ohio
Posts: 12,151


Oh my, isn't this an interesting can of worms. I think I will try to drag Denton back into it to make sure I haven't misrepresented what he said and to let him have his two cents worth in the matter.
__________________
Nick
__________________________
Orange Hat Family Member
CMP Certified GSM Master Instructor
NRA Certified Instructor
NRA Benefactor Member
"First contemplation of the problems of Interior Ballistics gives the impression that they should yield rather easily to relatively simple methods of analysis. Further study shows the subject to be of almost unbelievable complexity." Homer Powley

02222017, 01:41 PM

Beartooth Regular


Join Date: Mar 2014
Posts: 53


Leveraction66, In order to assist you with your OCW workup we need to see all the targets, preferably side by side, with the appropriate load information. What you are looking for is 2  3 groups that are centered on the same spot. This is a node. Alternatively, you can also locate the "scatter node" to help you find your node for loading. Typically, the nodes are 3% of charge weight apart. Therefore, from a scatter node, you can multiply or divide the charge weight for that load by 1.015 to find a good node area. You might want to check out ocwreloading.com for more information on the process. You can also post your result on Dan Newberry's forum, PracticalRifler.fr.yuku.com for more help/input/advice.
I usually shoot 3 shots per charge weight while working up my load. OnTarget is not really necessary for this. You can usually find the accuracy node and scatter node without it.
I usually will do a seating depth test prior to the OCW test to find my best seating depth. Some test for seating depth after locating the node.
If you need more help, let me know.
FWIW,
Dennis
__________________
Dennis
Neither common sense nor common courtesy are very common anymore!

02232017, 05:57 AM


Super Moderator


Join Date: Jan 2005
Location: Hilliard, Ohio
Posts: 12,151


Good advice from Dennis. The only caveat I'll toss in is that the last time I communicated with Chris Long by email (a decade ago, now, I think) he was talking about 2% between nodes for a particular gun. I gave that some thought and realized the percentage separation would vary with barrel length. This is because the bullet has it's greatest acceleration and velocity pickup near the pressure peak which is in the first few inches of the bore, so barrel time for lengths greater than that will be disproportionately shorter than the percent length added to a barrel, while the periodicity of the pressure wave is directly proportional to that barrel length. So the two don't track, and it is the relationship of these two that determine node spacing under Long's theory.
Based on my random target you can see that the center positions of three shot groups will wander some randomly. The population SD divided by the square root of the number of shots in the group gives the Standard Error (SE) which is just the SD of a population with an n times smaller variance. Averaging has that effect on variance, so you use the SE as the SD of the average location for groups of n shots. In the random group in the illustration, the population SD was defined as 1.0 moa when I generated it, so the standard error for three shots is close to 0.58 moa, and you can glean a sense of that magnitude of wandering from looking at the three green x's. You can also see that if you combined all nine shots, the resulting average location would be closer to the center where the the yellow cross indicating the true average for an infinitely large group (also defined when I generated the group) would fall. But if you average all nine shots in three shot groups by using a threegroup running average, as I described, you cut the error, displacement from the infinite shot group center, by 42%, which is significant improvement for eyeballing the thing to try to discern a spot.
F2G1D,
I'm still working on contact with Denton. In the meanwhile, here's a link to his article, The Perverse Nature of Standard Deviation, which I think is a pretty good attempt to explain SD to average shooters. His comment about n1 is in a footnote there. What he actually wrote, which didn't quite match my memory of it, is:
"Before 1925, everybody divided by n, but, then, {Fisher} came out with his book on ANOVA. n1 makes the math in ANOVA work cleanly, and the influence of ANOVA on the science of statistics was profound. After 1925, the world pretty much switched to n1. The choice of n or n1 is simply a matter of convenience, anyway, and does not spring from some great, secret truth, known only to statisticians."
Perhaps by making the math work "cleanly", he was referring to removing some degree of bias, and I realize I don't actually know. From the article I linked to, it clearly doesn't do a perfect job of that.
__________________
Nick
__________________________
Orange Hat Family Member
CMP Certified GSM Master Instructor
NRA Certified Instructor
NRA Benefactor Member
"First contemplation of the problems of Interior Ballistics gives the impression that they should yield rather easily to relatively simple methods of analysis. Further study shows the subject to be of almost unbelievable complexity." Homer Powley

Thread Tools 

Display Modes 
Rate This Thread 
Linear Mode


Posting Rules

You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off



All times are GMT 8. The time now is 04:09 PM.


