Armory Sponsor
Small Sample Testing (Page 2 of 2)
|
Performing Monte Carlo simulations is fine if you need to actually model the population distribution. For most reloading use cases, I think you'd be better off understanding which distribution represents the data of interest, and then computing the confidence intervals around the appropriate sample statistics. I actually started drafting a book on this, but I don't think it would be long enough to be a book. More like a couple dozen pages. One of these days I'll finish it. |
|
Originally Posted By Deimo5: Performing Monte Carlo simulations is fine if you need to actually model the population distribution. For most reloading use cases, I think you'd be better off understanding which distribution represents the data of interest, and then computing the confidence intervals around the appropriate sample statistics. I actually started drafting a book on this, but I don't think it would be long enough to be a book. More like a couple dozen pages. One of these days I'll finish it. That first step was to see and show just how far off the mean one might be, even with a small SD. Also, I wanted to see how many samples were needed to begin to actually see the probability distribution function (PDF) in the data. <-- It's a BIG number. The more I look into this, the less value I place on 3 and 5 shot samples. Oh well,... |
|
Originally Posted By Deimo5: Performing Monte Carlo simulations is fine if you need to actually model the population distribution. For most reloading use cases, I think you'd be better off understanding which distribution represents the data of interest, and then computing the confidence intervals around the appropriate sample statistics. I actually started drafting a book on this, but I don't think it would be long enough to be a book. More like a couple dozen pages. One of these days I'll finish it. I would encourage you to finish it and to release it, somehow, somewhere, because of all the misconceptions I have had about the stats shown on chrono screens and because of the chatter between shooters that I've read and heard. Perhaps a sticky thread summary on this forum would suffice with a link to a drive in the cloud. |
|
I have done some more work on this. I have a two dimensional shot distribution overlaid on an NRA target. This is useful for small sample sizes. I also have a shot distribution useful for larger sample sizes. This is also overlaid on an NRA target. They both assume the muzzle velocity is variable and the shooters's wind calls are imperfect (these are the input variables). I'm not sure where to go with it from here. I seem to have answered my original questions with the work I've already done. I need to develop a set of questions which can help direct my future efforts. Anybody and everybody, even a tyro, can shoot an occasional "X". Don't get too excited about it. <-- General Hatcher wrote something like this in his notes. |
|
Large sample sizes give you increased confidence in your hit potential due to accuracy. Large sample sizes also burn out barrels trying to find the right load. So, I build confidence using multiple medium samples and funneling techniques like OCW. You can also stop shooting when you blow a targeted group size to minimize shots with 10 shot medium samples. It’s a balance. |
|
Originally Posted By M15A2CLE: Large sample sizes give you increased confidence in your hit potential due to accuracy. Large sample sizes also burn out barrels trying to find the right load. So, I build confidence using multiple medium samples and funneling techniques like OCW. You can also stop shooting when you blow a targeted group size to minimize shots with 10 shot medium samples. It’s a balance. When I get a new-to-me cartridge, I also move from load development into load verification very quickly. In part, this is because of the barrel life issue but also because I am able to great advice from people on this Forum and fellow shooters at the Range. Typically, I use ~5 sets of 5 powder charges, then move to a 20 round verification of a single charge weight. After that the load is used for shooting from a bench or in a match. I carefully watch new loads but have never had a significant surprise with my reloads. |
| Your standard procedure is Interesting as I too often start with 5-shots, x 5 charge increments x 4-5 powders to eliminate the obvious... then shoot more groups... but instead of jumping to 20 shot groups, I continue with 5 but aggregate them later... but them I don't shoot any comps anymore- none around here I can find and no ranges longer than 100... yuk! |
|
Originally Posted By Trollslayer: That first step was to see and show just how far off the mean one might be, even with a small SD. Also, I wanted to see how many samples were needed to begin to actually see the PDF in the data. <-- It's a BIG number. The more I look into this, the less value I place on 3 and 5 shot samples. Oh well,... I've seen some discussion from Litz regarding small groups, including aggregate small groups. IIRC the conclusion was, small group sizes are good for making erroneous decisions. |
I know I'll never go home.
So set fire to your ships, and past regrets, and be free.
So set fire to your ships, and past regrets, and be free.
|
Originally Posted By SyberSniper: Your standard procedure is Interesting as I too often start with 5-shots, x 5 charge increments x 4-5 powders to eliminate the obvious... then shoot more groups... but instead of jumping to 20 shot groups, I continue with 5 but aggregate them later... but them I don't shoot any comps anymore- none around here I can find and no ranges longer than 100... yuk! My experience with people on line here and talking to other shooters, has meant I need only one or two powders. When l started loading 308, I needed to test only one powder - IMR4064. It came that highly rated by everyone. I still use it decades later. For 308, I may need to switch to another powder and it will likely be VARGET. People here online say it is very good in 308. That I already use it for my 223 is a bonus. |
Small Sample Testing (Page 2 of 2)
Armory Sponsor