Seating depth test

[Rickyd1]

Hi all, I’m building a load for my 6.5 creedmoor. I have selected the charge to be 43.7 n555. Gives a low e/s and s/d. I’m now going to do different seating depths on increments of 3thou. I have started at jam - 20 thou. How many steps should I do?

 

[Harman117]

0.003" ? Ridiculous increments. You'll be spending all your time testing if you go that small. I'd do more like 30 thou increments. And LARGE sample sizes. How did you determine your charge weight?

 

[Aftermath]

Hi all, I’m building a load for my 6.5 creedmoor. I have selected the charge to be 43.7 n555. Gives a low e/s and s/d. I’m now going to do different seating depths on increments of 3thou. I have started at jam - 20 thou. How many steps should I do?

What bullet?

 

[KYAggie]

Here’s some light reading that might help. I have not done this, but maybe someone else here has.

Bullet Jump Research: Executive Summary & Load Development Tips

This is the last post in a series of articles focused on bullet jump research that has been conduct...

precisionrifleblog.com

 

[Rickyd1]

140 eldm bullets. Determined powder charge by chronographing different charge weights. 43.7 gives e/s of 9.2 and s/d of 4.1 average speed is 2745. It’s a nice node as it sits between 2747 and 2744! Can’t get much better than that!

 

[AllenOne1]

Hi all, I’m building a load for my 6.5 creedmoor. I have selected the charge to be 43.7 n555. Gives a low e/s and s/d. I’m now going to do different seating depths on increments of 3thou. I have started at jam - 20 thou. How many steps should I do?

I would run 10 3 shot groups and see what you get. Anywhere that looks promising can then be tested with a larger sample set.

 

[Hickswr]

The component companies thank you for doing seating depth test in 0.003 increments. Don’t forget to account for the throat eroding that much over 200 or so rounds

 

[diggler1833]

I wouldn't do .003 jump testing. By the time you "found" what you liked, your throat would have eroded from other testing to the point where you were no longer jumping from the same spot.

I do like Mark's test in the link posted above. I personally do .015", and then occasionally will refine a little more if I have a barrel that seems more picky about seating depth on a particular bullet. I've tested windows in seating depth before, and found that it requires you to really stay on top of measuring and playing with your dies...all for a very minimal gain. I'm much more of a fan these days of trying another bullet until I get something that isn't so picky.

The .003 testing is much more popular with those trying to compete in Benchrest or F-Class. The beanbag and barricade crowd definitely isn't going to benefit at all from such small increments.

 

[KYAggie]

The component companies thank you for doing seating depth test in 0.003 increments. Don’t forget to account for the throat eroding that much over 200 or so rounds

Exactly. That PRB article is interesting. From reading it, it seems to me that there may be a seating depth for each different bullet where it’s most forgiving and that might not change with different guns? If that’s so, it would be awesome if it was published data for each different bullet as it would save us all a lot of time and money. Of course, the component companies wouldn’t want that…….

 

[Aftermath]

I like to do 10 shot groups. The incremental amount depends on which bullet. For instance, the Berger VLD's are known to like to be very close, very little jump. I use 0.003" jumps and have even done a set of 0.001" jumps with notable changes in group size. The Gamekings seem to prefer more jump so the increment is 0.010" and then testing in-between when I think i have found the spot. Once I do find a good enough depth, I don't chase the erosion. If accuracy really got real bad, I might, but I'd be more inclined to get a new barrel.

There are going to be folks on here saying that seating depth does not matter.

There will be those that try to tell you some statistics they do not understand means you have to shoot a proper sample size.

 

[RegionRat]

Here is the TLDR version from the link that @KYAggie posted above.

Here is an observation of my own in Mark's charted data. Notice how if you pick any one rifle and follow the line while watching the vertical scale, some guns are much tighter than others and stay that way for several steps of depths. Other guns jump up and down worse than the stock markets and it would be hard to pick a durable zone.

Based on the sum total of the chart, you might say that all of these guns shot pretty well and that you could just jump 0.020" and not worry about it at all. If anything, the context of a hunting or PRS rig where the ammo is all pre-loaded is one thing, a BR match where a relay is 5 targets of 5 shots plus sighters, but they can tune in real time between relays is very different than pre-loading.

There is nothing wrong with giving seating depth a try, other than the bbl life and resources spent on it. If the OP's rifle is for hunting or low volume shooting, then I would suggest he starts at jump 0.005" and takes it in steps of 0.005" to go seven or eight steps and see if it makes any real difference for him at all.

Here is an example of a match bbl in 6 Dasher with a Berger 105 Hybrid where the graph represents the vertical at 600 yards. Not counting the first two steps that were rediculously lucky where all three shots were touching at 600 yards, there are two places where the seating depth seems to be quiet. One zone is roughly 0.035" - 0.045" and the other at 0.060" to 0.080".

Keep in mind this is low sample testing of only three runs. Hard to say if these results would keep or if the verticals would spread out with more samples. The gun shoots just about anything into under 0.5 MOA at 600 for 20 shot strings. Bartlein bbl. chambered by Gary Eliseo.

 

[Dogtown]

mmmm....data

 

[CK1.0]

I find myself in the camp that doesn't think seating depth really matters. IMO, what matters is they all come out the same (all within a couple thou of each other when measured base-to-ogive).

I've tried many different methods and shot lots of rounds trying to discern if it makes a difference, for a long time believing the more jump the better (.080"-.100" off the lands), and then just as much time trying the other way (0.020" off or less)... In the end, the only thing I determined was that I'd be an idiot if I thought the monkey pulling the trigger was reliably consistent enough shot-to-shot to be sure my variations in seating depth were solely and irrefutably responsible for what I was seeing on targets.

Now I don't even measure or really care where my lands are, I just load all my rounds to the shortest length I can while keeping the bullet's bearing surface above the case's neck/shoulder junction (which is actually another thing I'm also not sure even matters), and my rounds always shoot (and always fit in my mags), every time, with every different barrel.

My average 100 yard group is .1 MIL/0.3MOA more or less on command... and whether I print a smaller or larger one probaly has more to do with how I slept the night before than my seating depth.

 

[Franko]

Over the years I have slowly come to the point where I think that seating depth is something that only benchrest shooters should care about. It is my opinion that rest of us simply don't have the gear, the skills, or the need to chase this particular dragon. I'd rather spend the money that it would take to perform real valid seating depth tests on each new barrel on better bullets, brass, and measuring tools.

 

[Seymour Fish]

0.003" ? Ridiculous increments. You'll be spending all your time testing if you go that small. I'd do more like 30 thou increments. And LARGE sample sizes. How did you determine your charge weight?

Cost of a few increments vs cost of Large , Absurdly Large, sample sizes ? Suggest only Hdy components to fatten your pocket

 

[Harman117]

Cost of a few increments vs cost of Large , Absurdly Large, sample sizes ? Suggest only Hdy components to fatten your pocket

I wish I was paid by Hornady, that would be sweet...

Are you okay man? This is the 3rd post of mine this morning that you're going back and commenting on

 

[Pharmseller]

Seating depth is crucial to me.

The only difference between these loads is the depth.

Then

Then

I usually do 4 iterations starting at the lands, just barely, then three 0.04” off.





P

 

[Pharmseller]

I have lots of examples like this, but this one is the most remarkable.




P

 

[CK1.0]

I hate to be “that guy” but 3 groups and 10 shots total isn’t data, it’s a Rorschach test.

 

[memilanuk]

I hate to be “that guy”

No, you don't.

 

[Seymour Fish]

No, you don't.

Monte , we have our own special version of Woke on exhibition . Trust the new religion

 

[JB.IC]

I’m surprised the data experts didn’t noticed a massive issue with the test design. I’ll leave that for them to find out.

From the data table they provided for test #5, we can see the vertical POI is associated with MV. There’s a likely reason the first few shots diverge and have nothing to do with bullet jump.

Because they failed to design the test properly, nothing of statistical value can be concluded. In fact, because of assumption violations, you can’t even use statistical tests.

 

[KYAggie]

I’m surprised the data experts didn’t noticed a massive issue with the test design. I’ll leave that for them to find out.

From the data table they provided for test #5, we can see the vertical POI is associated with MV. There’s a likely reason the first few shots diverge and have nothing to do with bullet jump.
View attachment 8539716
Because they failed to design the test properly, nothing of statistical value can be concluded. In fact, because of assumption violations, you can’t even use statistical tests.

Could be. How would you go about doing such a test in a way it actually could produce meaningful data? I have personally seen what I believe to be differences in precision based on different seating depths, but I’m still trying to figure out how to test in in a meaningful way. There’s so many different variables too. What works in one barrel may or may not work in another.

 

[CK1.0]

The problem with most of these “tests” is they automatically assume that the monkey pulling the trigger is infallible and ultra-consistent… and we all know that’s absolutely not the case.

For example, using the earlier post showing 3 groups with the last one clearly being the best, who’s to say the shooter just didn’t need a couple groups to warm up? That’s way more likely, yet because he was “testing” seating depth, he chose to ignore all the other more obvious possibilities that could be responsible and decided a few thou of seating depth difference was the culprit.

Its the Dunning-Kruger effect at work 99% of the time, guys overestimate their capabilities and then assume what they’re seeing is what they wanted to see.

If someone can lock their gun in a rigid mechanical rest that prevents the gun from moving and then activate the trigger using a device that can be 100% consistent then maybe they can convince me something like seating depth really matters. But at this point, much of the old “reloading lore” that gets thrown around as fact around here is equivalent to saying “You’ll shoot better when you wear a blue shirt, because I always shoot good when I wear a blue shirt”, just theory that’s closer to nonsense than anything factually provable.

Then someone will say “well benchresters and f-class so and so can’t be wrong” to which I call BS, because half of the bench rest guys I’ve seen are terrible shooters (even with the help of a 20lb $800 rest) and what discipline one shoots doesn’t make them immune to confirmation-bias, the placebo effect, and a long list of other fallacies.

 

[JB.IC]

Could be. How would you go about doing such a test in a way it actually could produce meaningful data? I have personally seen what I believe to be differences in precision based on different seating depths, but I’m still trying to figure out how to test in in a meaningful way. There’s so many different variables too. What works in one barrel may or may not work in another.

So there’s how it would be conducted in a lab, which I don’t think is useful to discuss here, and then there’s how you would do it in the field.

That being said, the most simplest designs are generally the best designs because they’re easy to implement and cheaper to conduct. Why? You can tweak pretty much everything on a weapon system. From the firing pin, bipod, and all the way down to how much you seat a primer. The combinations are massive. To test all of them, you’d need like 3 barrels and that’s a waste. So, some factors need to be held constant to make a test simple. For instance, only using one bullet, one powder charge, one primer type, one neck tension, and the same shooting environment, shooting surface, and shooting style (don’t switch from prone to bench in the middle of a test).

If I was curious about bullet jump variation and wanted to satisfy statistical assumptions, I would choose two or three bullet jump settings that are far a part and have multiple observations per setting. I’d decide what settings would be practical. For instance, I do not like to jam rounds. So I wouldn’t test it. Something like 0.02”, 0.08”, and 0.14” jump. Then I would shoot at least say 20 rounds per jump. All in random order under the same firing conditions; 60 rounds total, all in one shooting session, annd all in random order.

Variance is additive. Variance of the shooter, variance of the rifle at a particular ammunition config, variance of the ammo manufacture (you), variance of the ammo components, and variance of the environment all sum together. Trying to partition variance and assigning it to a difference in bullet jump of say 0.005” is a pipe dream. Not in most field conditions will anyone be able to do that. There are some good bench rest and f-class shooters that deserve a little more trust in their methods, but I’ve examined some of their test designs and they’re not that great. I’ll say this, they’d never be published in a scientific journal or any DoD research lab wouldn’t take their results seriously, but people will eat it up on YouTube.

 

[JB.IC]

The problem with most of these “tests” is they automatically assume that the monkey pulling the trigger is infallible and ultra-consistent… and we all know that’s absolutely not the case.

For example, using the earlier post showing 3 groups with the last one clearly being the best, who’s to say the shooter just didn’t need a couple groups to warm up? That’s way more likely, yet because he was “testing” seating depth, he chose to ignore all the other more obvious possibilities that could be responsible and decided a few thou of seating depth difference was the culprit.

Its the Dunning-Kruger effect at work 99% of the time, guys overestimate their capabilities and then assume what they’re seeing is what they wanted to see.

If someone can lock their gun in a rigid mechanical rest that prevents the gun from moving and then activate the trigger using a device that can be 100% consistent then maybe they can convince me something like seating depth really matters. But at this point, much of the old “reloading lore” that gets thrown around as fact around here is equivalent to saying “You’ll shoot better when you wear a blue shirt, because I always shoot good when I wear a blue shirt”, just theory that’s closer to nonsense than anything factually provable.

Then someone will say “well benchresters and f-class so and so can’t be wrong” to which I call BS, because half of the bench rest guys I’ve seen are terrible shooters (even with the help of a 20lb $800 rest) and what discipline one shoots doesn’t make them immune to confirmation-bias, the placebo effect, and a long list of other fallacies.

My response touches on some of the stuff you mentioned. I generally don’t like to start beef with people in our supposed community because we are all we have at the end of the day. But there are a lot of people overstating their knowledge on here, YouTube, or some blog. I mean just the other day I deleted all my comments out of a thread because the OP didn’t know what a key concept about a distribution and I realized maybe I shouldn’t go into any depth about the subject until they take a probability course.

 

[Aftermath]

So much non-sense.

How do you determine a statistically valid zero? 20 or a hundred rounds?

I tell you what I do and you are going to shit your pants. I look through the scope and aim at where I want the bullet to impact. I squeeze the trigger. I observe that the bullet did not impact where I wanted it to.

Now, get this. (Grab the toilet paper)

Rather than shoot a hundred more times to make sure it's not the fucking trigger monkey, I go ahead and adjust my scope.

How many shots do you take to come up with a statistically viable SD? How many times should one measure the powder charge, the CBTO, the primer seating depth....in order to make sure it is statistically viable and not the scale/caliper/dial monkey?

 

[Aftermath]

Just think....if I were to make a bakers dozen shit cookies, some here would take more than the first bite.

 

[Pharmseller]

Just think....if I were to make a bakers dozen shit cookies, some here would take more than the first bite.

What kind of shìt?




P

 

[johnjohn301]

With my own little knowledge….. For your load development with the 6.5 Creedmoor, if you're starting at jam - 20 thousandths and planning to adjust your seating depth in increments of 3 thousandths, you can determine the number of steps based on how deep you want to go.
Typically, you might want to test a range from jam - 20 thousandths to jam + 20 thousandths, which gives you a total of 40 thousandths to cover. To find out how many steps that is, you can divide the total range by your increment: 40 thou / 3 thou = approximately 13.33 steps. Since you can't have a fraction of a step, you would round that to 13 steps. So, you could do 14 total positions, starting from jam - 20 thou up to jam + 20 thou.

Make sure you keep track of your results for each seating depth, and you should be good to go!

 

[CK1.0]

Just think....if I were to make a bakers dozen shit cookies, some here would take more than the first bite.

It's actually not like that at all.

If one chooses to view reloading and load development as a religion, then I guess they can believe what they want to believe, but 99% of the fancy cardboard "tests" guys post trying to show "nodes", ladders showing "flat spots" and other related nonsense is cringy as hell and it just becomes sad at some point.

It must hurt when someone who believes they possess some sort of knowledge is told or finds out that they don't really know shit... but if they cannot show anything statistically relevant to back it up, or refuse to even try, then they should shut up and stop spreading their BS to new shooters.

Some of the guys on here have been trading in the same "reloading lore" BS for so long that they've convinced themselves it's fact when it's not, period, with many of them having taken the stance of "if I'm wrong, I don't want to be right".

When newbies post about nodes and other known nonsense, the correct answer post should be "stop that shit", anything else is unproductive and just perpetuating the nonsense, "that's the way I've always done it" isn't a real reason for doing anything.

 

[Aftermath]

It's actually not like that at all.

If one chooses to view reloading and load development as a religion, then I guess they can believe what they want to believe, but 99% of the fancy cardboard "tests" guys post trying to show "nodes", ladders showing "flat spots" and other related nonsense is cringy as hell and it just becomes sad at some point.

It must hurt when someone who believes they possess some sort of knowledge is told or finds out that they don't really know shit... but if they cannot show anything statistically relevant to back it up, or refuse to even try, then they should shut up and stop spreading their BS to new shooters.

Some of the guys on here have been trading in the same "reloading lore" BS for so long that they've convinced themselves it's fact when it's not, period, with many of them having taken the stance of "if I'm wrong, I don't want to be right".

When newbies post about nodes and other known nonsense, the correct answer post should be "stop that shit", anything else is unproductive and just perpetuating the nonsense, "that's the way I've always done it" isn't a real reason for doing anything.

So...second bite? Third? How many turds would you eat before it was "statistically relevant"? You have a finite set. How many would be enough? What would be an acceptable minimum sample set?

Back to the SD and seating depth and neck tension and powder charge and...and...and...
How many measurements are good enough for you to believe you have achieved statistical relevance?

How can you be sure any error was not the monkey on the dial caliper creating those minuscule discrepancies?

I can be statistically POSITIVE, with 100% correlation, infinitely high statistical significance, with one single shot fired. Hundreds of thousands of other shooters can be, as well.

Same with a shit cookie.

 

[CK1.0]

So...second bite? Third? How many turds would you eat before it was "statistically relevant"? You have a finite set. How many would be enough? What would be an acceptable minimum sample set?

Back to the SD and seating depth and neck tension and powder charge and...and...and...
How many measurements are good enough for you to believe you have achieved statistical relevance?

How can you be sure any error was not the monkey on the dial caliper creating those minuscule discrepancies?

I can be statistically POSITIVE, with 100% correlation, infinitely high statistical significance, with one single shot fired. Hundreds of thousands of other shooters can be, as well.

Same with a shit cookie.

I'd eat zero turds because that has nothing to do with this and most of us can tell that you're on your heels (what you're employing is called the loaded question fallacy... usually one of the first tactics someone on the side of a losing argument reaches for when they don't have anything valid to argue).

It might be a surprise to some, but this shit actually isn't that hard, if one can be the best ammo factory they can be and turn out ammo where every round they produce comes out nearly the same, with SDs consistently in the single-digits over 20 shots, their groups will be small and their long-range waterline will be fine. No magic recipes or reloading lore myths necessary.

 

[Aftermath]

I'd eat zero turds because that has nothing to do with this and most of us can tell that what you're employing is called the loaded question fallacy... usually one of the first tactics someone on the side of a losing argument reaches for when they don't have anything valid to argue.

It might be a surprise to some, but this shit actually isn't that hard, if one can be the best ammo factory they can be and turn out ammo where every round they produce comes out nearly the same, with SDs consistently in the single-digits over 20 shots, their groups will be small and their long-range waterline will be fine. No magic recipes or reloading lore myths necessary.

So...being statistically relevant is only important when you say so. I see you are unable to answer simple questions with relatively simple mathematical (statistics!!) answers.

99% of the fancy cardboard "tests"....single-digits over 20 shots
Again...where do you get your numbers?

 

[Aftermath]

I'd eat zero turds because that has nothing to do with this and most of us can tell that you're on your heels (what you're employing is called the loaded question fallacy... usually one of the first tactics someone on the side of a losing argument reaches for when they don't have anything valid to argue).

It might be a surprise to some, but this shit actually isn't that hard, if one can be the best ammo factory they can be and turn out ammo where every round they produce comes out nearly the same, with SDs consistently in the single-digits over 20 shots, their groups will be small and their long-range waterline will be fine. No magic recipes or reloading lore myths necessary.

When does the trigger monkey stop being relevant during this non-magical shit that actually isn't that hard where there are small groups and flat waterline?

 

[CK1.0]

So...being statistically relevant is only important when you say so. I see you are unable to answer simple questions with relatively simple mathematical (statistics!!) answers.

99% of the fancy cardboard "tests"....single-digits over 20 shots
Again...where do you get your numbers?

The number isn't really important (though, a statistician, which I am not, might tell you that you either need 100 or at least 30 in your sample size to have something statistically relevant).

What is important is the "why".

Why bother?

When you miss or shoot a bad group, do you really believe it's the gun and not you?

Do you really believe you're such a skilled shooter that differences of a few thou in seating depth or a few tenths of a grain of powder can have more effect on your targets than the myriad of other more probable possibilities?

Do you always believe everything you think?

 

[Aftermath]

I'd eat zero turds because that has nothing to do with this and most of us can tell that you're on your heels (what you're employing is called the loaded question fallacy... usually one of the first tactics someone on the side of a losing argument reaches for when they don't have anything valid to argue).

It might be a surprise to some, but this shit actually isn't that hard, if one can be the best ammo factory they can be and turn out ammo where every round they produce comes out nearly the same, with SDs consistently in the single-digits over 20 shots, their groups will be small and their long-range waterline will be fine. No magic recipes or reloading lore myths necessary.

What you claim as "loaded question fallacy" shows your lack of depth.

Would you agree that a very small sample of one bite of one cookie (less than an entire cookie) would lead you to believe, right or wrong, that the rest of that cookie, along with the entire bakers dozen cookies, are shit?

You see, even statistics is contextual. An extremely small sample size can produce extremely reliable correlation. A huge sample might create an indeterminate.

However, if you want to live by the science, then fucking live by the science. Use good methodology to determine how many samples makes for a statistical correlation. To pick some random number because you heard that from a couple of guys that seemed smart is stupidity.

Don't pick and choose when and where you want to apply science and call yourself holy.

Do you always believe everything you think?

HAHAHAHA!!!

 

[CK1.0]

When does the trigger monkey stop being relevant during this non-magical shit that actually isn't that hard where there are small groups and flat waterline?

Never, that's the point.

With consistent rounds that produce a low SD, I can shoot a 0.1mil group with 32gn at .020" off, 34gn at .100" off, 33.3gn at .060" off, the load recipe doesn't matter, what matters is that they're all the same.

 

[JB.IC]

Assuming MV and group dispersion follows the normal distribution, which it does in a local sense:

To capture the expectation of the distribution 20 shots, on average, does a nice job

To properly capture the variance, 20 shots, on average, does a decent job, but a little more would be ideal. Variance is a lot harder to quantify than the expectation. So, group dispersion and MV dispersion, on average, are underestimated with small sample sizes.

A glimpse of the variation of the variation when it is not aggregated

 

[CK1.0]

What you claim as "loaded question fallacy" shows your lack of depth.

Would you agree that a very small sample of one bite of one cookie (less than an entire cookie) would lead you to believe, right or wrong, that the rest of that cookie, along with the entire bakers dozen cookies, are shit?

You see, even statistics is contextual. An extremely small sample size can produce extremely reliable correlation. A huge sample might create an indeterminate.

However, if you want to live by the science, then fucking live by the science. Use good methodology to determine how many samples makes for a statistical correlation. To pick some random number because you heard that from a couple of guys that seemed smart is stupidity.

Don't pick and choose when and where you want to apply science and call yourself holy.


HAHAHAHA!!!

No dude, your whole thing is based on confirmation bias.

If I took a single shot at a target and it was 3" high and 2" left of my POA and normal POI, I would know nothing (and would look in the mirror and shoot some more, before declaring unequivocally that the load was shit).

 

[Aftermath]

No dude, you're whole thing is based on confirmation bias.

If I took a single shot at a target and it was 3" high and 2" left of my POA and normal POI, I would know nothing, and would look in the mirror before declaring that load shit.

I see.
So you would shoot some more, maybe some minimal sample size as determined using sound statistical methods, maybe based upon expected barrel life, or maybe calculated rounds available with the components on hand, to verify that the fault was the monkey. I can tell you that I know a large number of shooters, and I will include myself, that would be able to say that the shot was released clean and make adjustments rather that waste time and commodities verifying or disproving some statistical bias. I mean, you wouldn't want to just pull some number out your ass, like 20, and say that's good. Right?

Never, that's the point.

With consistent rounds that produce a low SD, I can shoot a 0.1mil group with 32gn at .020" off, 34gn at .100" off, 33.3gn at .060" off, the load recipe doesn't matter, what matters is that they're all the same.

Exactly. The trigger monkey is never NOT part of the equation but this is the very first time you have admitted that, possibly in error of what you mean. (Earlier in this thread and in others, you propose a rifle with a remote trigger and shit.) The rifle, cartridge, shooter, etc is a system. Take a part of the system away, add a piece to that system, you need to now retest everything all over again if you want to live by that science.

 

[Aftermath]

Assuming MV and group dispersion follows the normal distribution, which it does in a local sense:

To capture the expectation of the distribution 20 shots, on average, does a nice job
View attachment 8540007

To properly capture the variance, 20 shots, on average, does a decent job, but a little more would be ideal. Variance is a lot harder to quantify than the expectation. So, group dispersion and MV dispersion, on average, are underestimated with small sample sizes.

View attachment 8540009

Nice little charts.
Why does the green/yellow chart only go to 30 observations?
Isn't there a more definitive method to determine minimum sample size that an actual statistician would use?

 

[JB.IC]

Nice little charts.
Why does the green/yellow chart only go to 30 observations?
Isn't there a more definitive method to determine minimum sample size that an actual statistician would use?

It only goes to 30 because it never converges to the true population standard deviation. You can put the sample size to 100000000 and it won’t converge. Standard deviation is a biased estimator. I like how you’re claiming I’m not a statistician lol. I guess my degrees and time at research labs was for nothing.

 

[Aftermath]

It only goes to 30 because it never converges to the true population standard deviation. You can put the sample size to 100000000 and it won’t converge. Standard deviation is a biased estimator. I like how you’re claiming I’m not a statistician lol. I guess my degrees and time at research labs was for nothing.

I see you feel some sort of attack.
There was not any implication to impugn your degrees or whatever.
I, too, have degrees...note the user name.

So...let's use expected barrel life of some caliber as 5000 rounds. A statistically significant correlation of thousands of rifles of that caliber over an infinite period of time has shown the barrels are shot out at 5000 rounds. Our maximum set is 5000 shots.
How many rounds should be fired (sampled) to determine with a 99% confidence level that the SD of the MV is within 1% of the measured value?

EDITED for some grammar shit that was driving me crazy

 

[JB.IC]

I see you feel some sort of attack.
There was not any implication of impunity on your degrees or whatever.
I, too, have degrees...note the user name.

So...let's use expected barrel life of some caliber as 5000 rounds. A statistically significant correlation of thousands of rifles of that caliber over an infinite period of time has shown the barrels are shot out at 5000 rounds. Our maximum set is 5000 shots.
How many rounds should be fired (sampled) to determine with a 99% confidence level that the SD of the MV is within 1% of the measured value?

You’re using words that I don’t think you know what they mean. Or whatever.

You’re welcome to solve for n if you want.

It’s a nice game you got going on here

 

[Aftermath]

You’re using words that I don’t think you know what they mean. Or whatever.

You’re welcome to solve for n if you want.
View attachment 8540091

It’s a nice game you got going on here

I see.
You choose to impugn my claims of degrees without any effort to solve the straightforward stats problem.
With your degrees and years in the labs and shit, I should think it would be a simple task.
Even HS students can solve that.
A rudimentary search for a stats calculator and you plugging in the numbers in the correct slots would yield you a valid solution.
A fucking moron could find that solution.

I believe you choose to leave that question unanswered because a correct answer would invalidate all of the nonsensical "statistical relevancy" claims made above.

 

[JB.IC]

I see.
You choose to impugn my claims of degrees without any effort to solve the straightforward stats problem.
With your degrees and years in the labs and shit, I should think it would be a simple task.
Even HS students can solve that.
A rudimentary search for a stats calculator and you plugging in the numbers in the correct slots would yield you a valid solution.
A fucking moron could find that solution.

I believe you choose to leave that question unanswered because a correct answer would invalidate all of the nonsensical "statistical relevancy" claims made above.

Show me where I called into question your degrees specifically.

The plots I provided are based on statistical theory and their distribution. Nothing to do with my opinion or claims.

Unless I read your question incorrectly, there is no closed form solution for the sample size needed to ensure the 99% confidence interval width of the SD is no larger than 1% of the calculated SD. That’s a numerical method for solving.

Did you mean to ask how large of a sample is needed so that the margin of error of the MEAN is 1% given a population of 5000? If that’s your question, than it’s around 3000. There is a closed form solution for that but it’s a different question than you asked.

I don’t understand why you’re coming at me when I have yet to be a dick to you. Hell I agree with some of the sentiment you’re saying. All I did was post statistical theory.

 

[Pharmseller]

You fellas shur do talk purty.





P

 

[Pharmseller]

Impugn and rudimentary.

Makes me hot.




P

 

[Aftermath]

Show me where I called into question your degrees specifically.

You’re using words that I don’t think you know what they mean. Or whatever.

The plots I provided are based on statistical theory and their distribution. Nothing to do with my opinion or claims.

I understand where the plots and charts came from. I am sure I can find similar plots and charts in several books in the storage unit.

Unless I read your question incorrectly, there is no closed form solution for the sample size needed to ensure the 99% confidence interval width of the SD is no larger than 1% of the calculated SD. That’s a numerical method for solving.

Did you mean to ask how large of a sample is needed so that the margin of error of the MEAN is 1% given a population of 5000? If that’s your question, than it’s around 3000. There is a closed form solution for that but it’s a different question than you asked.

You are correct, I should have referenced the mean of the SD of the MV as measured. Your estimate of 3000 is not so low. It's closer to 3800 shots taken to determine the validity of an SD. Not 20, not 30.....Three thousand eight hundred rounds fired to statistically determine the validity (or not) of that measurement.

You only have 1200 rounds to shoot before you start the process all over again.

To pick and choose when one wants to apply statistics while claiming someone else's number they chose is not valid just doesn't hold water.

If one decided that 20 was the magic number to call SD good, ok. I don't give a shit. But to say that a guy choosing to make a scope adjustment after a single shot is heresy; to say that 3, 5, or whatever number of shot groups does not stand up to the "Statistics", that argument just doesn't hold good water.

Show me where I called into question your degrees specifically.

I don’t understand why you’re coming at me when I have yet to be a dick to you. Hell I agree with some of the sentiment you’re saying. All I did was post statistical theory.

You have thin skin.