Anyone ever use the 2-shot-group / t-test method?

[blockfort]

I found this while looking for (overly technical) ways to analyze my groups.

http://www.public.iastate.edu/~jessie/PPB/Stats/Testing%20loads.htm

[bubbala]

So,,, did he ever figure out how many angels can fit on the head of a pin?

[Whiterabbit]

Love it. Bookmarked it.

There's only one thing I disagree with. His assertion that the "Best way" is tedious. Wrong!

We have a great free tool called OnTarget that will take a photographic image of your target and spit out all the raw data you need to perform the calculations.

Then you spend about an hour and a half the first time setting up excel with all the equations he helpfully provides. Drag and fill the chart and you have all the statistics he recommends you pull, Scroll to the bottom and you have your answer.

Next set of targets, the work is done. You just export to excel from OnTarget then the equations self update and you have your new answer.

It's brilliant, I'm glad he posted how to go from X-bar and Y-bar to t.

(BTW I think the 2 shot idea is brilliant too. But "the best way" is just so easy if you download OnTarget and have excel. May as well use that)

[noylj]

Don't forget:
The shooter can not know what ammunition is being fired and all data sets are random (you can't shoot all xyz ammunition, followed by abc ammunition, and then def ammunition)
And
The gun must be in the SAME condition (at minimum, same cleanliness and temperature) for each group
There may be some other requirements, but those are the two that leap out at me

[blockfort]

Don't forget:
The shooter can not know what ammunition is being fired and all data sets are random (you can't shoot all xyz ammunition, followed by abc ammunition, and then def ammunition)
And
The gun must be in the SAME condition (at minimum, same cleanliness and temperature) for each group
There may be some other requirements, but those are the two that leap out at me

This can be mitigated by shooting round robin, so that the ambient and gun conditions affect all ammo equally. So you can shoot X, A, D, then X, A, D again, etc etc, so as the weather and barrel heat up, all ammo is affected the same.

[blockfort]

And What about STDEV vs CEP (circular error probable), does it really matter which we use to characterize our results, they can "easily" be converted back and forth between each other. I say "easily" in quotes because I've found no less than 8 methods for calculating CEP, and they all have more or less error depending on the conditions of the experiment.

One good paper on CEP

Another about CEP methods compared.

[cindynles]

T-Test only tells you if there is a statistical difference between 2 groups (populations).

If you really want to dig into this, take a couple of classes to learn about D.O.E. (design of experiments). With D.O.E. you can build a mathematical model that will tell you what happens when you adjust parameters. For example if you run a D.O.E. and one of your factors is OAL, the model will predict what you output (say velocity) will be if you change it.

The nice thing is that you only need to test 2 or 3 points (hi and low, or hi, med, low). The model will also show you if there are interactions between different variables (for example powder and primer combinations).

[Whiterabbit]

heh. I am looking forward to experimenting with that, but keep in mind, the DOE model of just high and low with a couple variables (IE seating depth and powder charge) alone will soak up between 48 and 64 rounds if not closer to 100 after all the permutations and multiplier (to get group characteristics and SD, each of which are just one data point)

I do like the website recommendations. And the T-test complete with limitations is a GREAT idea! knowing that two populations are statistically the same can save you a huge number of components when you end up with a 2" group and a 1" group that are within the same population.

[cindynles]

heh. I am looking forward to experimenting with that, but keep in mind, the DOE model of just high and low with a couple variables (IE seating depth and powder charge) alone will soak up between 48 and 64 rounds if not closer to 100 after all the permutations and multiplier (to get group characteristics and SD, each of which are just one data point)

I do like the website recommendations. And the T-test complete with limitations is a GREAT idea! knowing that two populations are statistically the same can save you a huge number of components when you end up with a 2" group and a 1" group that are within the same population.

Only if you plan on performing a full factorial. Typically you perform a screening DOE (Taguchi, fractional factorial, or Latin) to identify the significant factors. Then you perform your modeling DOE on those factors.

The best part is that once you've completed the modeling DOE you can use the data to predict expected results. I've used this several times to speed up load development.

[Whiterabbit]

Yes, full factorial. High/low only, two factors of charge and seat depth. With a possible third discrete input of primer choice to compare my everyday primer with BR.

Ends up needing something like 8 datapoints minimum, recommended 12. Or something similar. Figure on either 5 shot groups or maybe 4 shot groups and double the testing in case a shot is bed, and you are looking at 50-100 rounds.

[MrOrange]

Some time ago, I read of a statistical attempt to come up with the minimum number of rounds in a group that could be used as an accurate representation of the precision capabilities of the firearm, and after much maths the scientific types came up with: 7. That's my answer, and I'm sticking with it.

As a practical matter, I often confirm a zero on a familiar firearm with just one shot. Since I know what group size the firearm is capable of, if that one hit would be anywhere within the nominal group size, I'm good.

On a side note, any casual shooter who's tried the American Rifleman's standard protocol of five consecutive five-shot groups knows that it is a tough challenge just to maintain concentration over that stretch, let alone continue to shoot at your top level. Something to consider when you look at their gun tests.

[bsumoba]

My buddy and I were speaking to a well respected person in the industry who does bullet testing for a big name bullet mfg and he said something that was quite interesting, and I am paraphrasing as the discussion was lengthy...

If you cannot, at 100 yards, get 2 shots to touch or go in one hole, why spend more time on that load?

This assumes a few things, you are using great equipment (not factory stuff), you know how to load good ammo, you are in controlled conditions (which he does, as he shoots in a tunnel), and you have a good set up (rest, rear bag, etc).

That being said, we cannot all shoot in a wind tunnel, but more times than not, if I see a group where the bullets do not all touch, I am typically moving on to the next load. When I finish developing a load, 5-shot groups will all touch, if not group in the 1's and 2's at 100 yards and it can do sub 1/2 MOA for 10+ shots at 100 yards.

Another well respected shooter once said and again I am paraphrasing, "The barrel is going downhill the moment you shoot the first round out of it"

Point being (as I interpret it), you do not want to put more rounds down the barrel than needed, especially if it is a good barrel. It is depressing when you have an awesome barrel and you keep testing and testing to the point you shoot the life out of the barrel. Save the good barrels for big matches or for times when you need it, and shoot the others for practice. Barrel heat is what kills barrels. So, when doing your load development, don't test every bullet/powder/primer combo. Do your research, stick with a bullet you will normally shoot, a known powder for that cartridge and bullet weight and work up a good load.

Most F-Class competitors will retire barrels from big matches in less than 1,000 yards and benchrest shooters might even retire them in fewer rounds. Tactical rifles do not necessarily have/need the accuracy requirements compared to F-Class or Benchrest, so they can go further because 1/2-5/8 MOA is still competitive when shooting steel targets.

More often than not, I am either finding a good load within 50-100 shots, or I note it as an "okay" barrel and simply shoot those barrels for practice matches monthly matches.