Rod bend vs rod unload

[VGB]

Yep, but the premise being tested was " The stiffest rods will create the longest casts!". The variance is casters prevented the premise being tested.

[Walter]

So they disproved the original hypothesis.

[VGB]

Yes, as you would expect. Caster ability and physiology being potential points of difference which takes us neatly back to Post 1 in this thread and Ron's assertion.

[Lasse Karlsson]

I have a suspicion that the underlying reason for the test, was the disqualification of a norwegian caster, that won their distance event with a very stiff rod made especially for throwing the XXD 5 weight line far. He was disqualified because the rod hadn't been available in shops the required amount of time (I think it was 2 weeks?) So his competitors hadn't been able to buy the wonder rod of slinging the line far. The test neatly shows there is no free lunch IMO, the rod in question is included in the 8 rods 8 casters.

Cheers
Lasse

[Walter]

I was told that beer was also involved but they waited until after the test so it wouldn’t affect the results.

They kept the expense down by not including Paul.

[VGB]

I wouldn't take part in a trial unless beer was involved

[Walter]

Getting ferociously expensive again…

[VGB]

I once did some trials planning in Munich that was so detailed that we couldn’t understand the script the next morning

[Torsten]

Hi Walter,

One of the confusions around the hypothetical hypotheses floating around is that we can get an absolute value of how far a caster can cast with any given rod if we just eliminate enough variables and do enough casts. Maybe even get a casting machine.

I think it's plausible that the casters adapt their technique to the gear and that's difficult to simulate with a casting machine.

But standard statistics tells us that for a normal curve 99.7% of casts we measure will be within 3 standard deviations of the mean.

3x StdDev means 6m and that's not accurate enough, at least if we also consider tournament casting. There already 1-2m make a large difference in my opinion. I'm looking myself for the right metric, maybe we need to ask a statistics prof for our problem. I think we want to compare also two overlapping distributions.

Of course the results will have a wider spread than if we had the true values for mean and variance but we still would have to show that it’s a true normal curve, which it isn’t actually.

That's indeed another issue, we don't know the shape of the distribution.

Taking that a step farther, if we take a 1 caster 2 rod test how many casts do we need to take with each rod to determine that there is a significant difference for that one caster?

Something like the 8x8 study is in my opinion currently out of reach (at least indoors), the math required isn't really impressive, just organizing such an event is not that easy. But 1 caster, 2 or 3 rods, outdoors should be possible.

Anyway, I've looked around for software and found G*Power,
https://www.psychologie.hhu.de/arbeitsg ... gie/gpower
that seems to be popular for social/biomedical research and maybe useful for us too.

Possible hypotheses are:
H0: There is no difference between the maximum casting distance given two rods A and B with a different stiffness for one caster.
H1: There is a difference - ""

I've tried the following settings:
* Means: Difference between two independent means (two groups)
* A priori: Compute required sample size - given alpha, power, and effect size
* Tails(s): Two
* Effect size d: 1 (I've assumed that we might see a difference of 2m and a StdDev of 2m)
* alpha: 0.05
* Power: 0.8 (that's a popular choice)
* Allocation ratio: 1

Result: Sample size per group: 17 / Total 34

The required sample size depends on the difference, that we want to detect, for only 1m I'd need 64 samples/group with the above settings.

Greetings,
Torsten

[Walter]

Thanks Torsten, that’s good stuff.

I was thinking it could be close enough to normal with both positive skew and kurtosis but a quick search tells me it may be a negative skew with positive 4th moment. Definitely those values will not be zero.

I will see if I can find something a bit more definitive.