I'm confused by what you are describing.

Does the chronograph capture multiple velocities for the same shot?

Yes

Labradar would catch 20-50 velocities, as the bullet goes downrange.  And then extrapolate back to its face, to give you "muzzle velocity".  It's how the unit tech works.

And yes, I think Garmins approach is more statistically correct, because it includes the error of every shot. They wont let you get and see that data, only the shots extrapolated MV value, so you can't repeat that Calc, but their SD includes the data error in it - or so it seems.

What is encouraging is this implies ot IS internally storing the trace data.  Meaning dome day we may learn how to get that (so we can calculate BC's)

The sample Standard Deviation (SD) would be more appropriate than the population SD, but it's not a huge factor for me.  I tend to look more at the extreme spread (max velocity - min velocity), then secondarily the SD.  As long as you are consistent in which is used for comparison of different loads/shot strings (population SD vs sample SD), it's a relative gauge of "Goodness".  The smaller the better.

Thanks, I have an 35P & Magnetospeed. I guess you could say that the Oehler captures multiple velocities with the proof screen, but I have not used that thing in forever since I got the Magneto speed. I just couldn't get my head around the population SD.

Doesn't matter because in the end, the average velocity is what's used as the input variable in a ballistics solver. From an academic perspective, I agree with your reasoning, sample is the one to use. The difference in the end is negligible as long as the numbers don't get mixed up on the comparison (ie. Sample compared against sample, population compared against population). There was a similar discussion about this on another forum.

Average and ES is really what matters. SD certainly shows how well you can load consistent ammo. I'd consider deleting a shot or two from my string if they skew the avg too much.

JMO, average is important but ES is mostly irrelevant.  SD, derived from a reasonable sample size is more important.  

ES derives from only for the sample you measure.  

SD, with some analysis, tells you what you can expect when you shoot more of that load.  It should be used to set your expectations.  

ES should not be used to set expectation because you will see greater variation than your current ES value as you shoot more rounds.



What seems to be missing from all the discussions of statistics is discussion of selecting an appropriate sample size to measure.  

As if to say, "I'm going to the range today.  How many rounds do I need to shoot to get a good idea of this load's performance?"

I mean typically I verify at 1007yds. My goal is to have less vertical than .2mil closer to .1mil. ES in the teens or even low 20s achieves this and results in single digit SDs. Be it 3 or 8 doesn't make much difference and I can't remember a time when I had a low ES and the SD wasn't low as well.

My 02... I haven't looked into the differences in how they are calculated, but I've noticed the unit shows an SD usually 1-2 fps or so lower than than the Excel function run on the same data set.  Sample sizes are normally 10-20 rounds.
I doubt it will ever be any kind of significant issue for me... seems like I'll just be looking for the lowest number for any given load development, so as long as it's a consistent comparator, I think its going to be fine.

Worst case, export the data and calculate it according to your preferred method.

When exporting from my CED it has an option to use sample (n-1) instead of population (n) in the formula. I always selected it for 5~20 shot groups. In Excel, the Stdev function produces the sample version by default- so I ignore the garmin SD and cacl my own from the individual shot velocity.
I think sample is the correct version because:  
All those environmental conditions and shooter readiness factors will always be different - for every load you ever shoot and can't be controlled, what you are trying to evaluate is how good a particular load is vs another one.

I may load up 500 rounds (my populatioin) but only chrono 5 in rifle A today, 10 in rifle B tomorrow and 5 in rifle A again - so only 20 of 500 get evaluated (my samples). Then next year I can periodically shoot a few more (samples) and add to my numbers to where I get a pretty good idea where that load performs.

I did a little research on SD /ES when setting up my Excel spreadsheet - my take on it is that although 5 shots can give you an idea, that 20 should be the minimum to start forming an opinion. I tend to group multiple sessions of 5-shot results of the same load in the same gun for better averaging of data sets- I think the sample method is more correct for this application.  

One can also calculate MAD (Mean Average Deviation) and RSD% (Relative Standard Deviation %)
- MAD is a measure of statistical dispersion, being more resilient to outliers in a data set than the standard deviation. With MAD, the deviations of a small number of outliers are irrelevant (Hi/Lo fps anomolies). In the Standard Deviation, the distances from the mean are squared, so large deviations are weighted more heavily, and thus outliers can heavily influence it.
- The % RSD value is based on the spread of results compared to the average. The result is expressed as a percentage - A lower percent  means that the spread of the results is small while a higher value indicates a larger spread of results.

Really it all academic in the end, just something to look at and say yep, its good or naw, keep looking...