 The Clarinet BBoard
|
Author: spikey1973
Date: 2023-10-23 19:22
Hey all,
Just a thought I am expressing here, so I am not stating this as right or wrong in anyway. So please everybody make up your own mind but if things below could be helpfull to you.
But since I am once again testing and trying things out with MP's, reed, ligatures en embochures and stuff (besides excersize ofcourse) as always using the single variable design approach to my testes. Basically because I didn't know better, or even aware of other working approaches, untill...
I bumped into this fairly unrelated youtube movie:
https://www.youtube.com/watch?v=5oULEuOoRd0&t=881s
explaining (IMHO) a fairly efficient way to test things with a multivariate approach. The only thing that is really needed (and I feel that everyone should figure this out for themself) that every variate needs to be able to be noted down using simple criteria. But if one can, then it might improve efficiency of finding your way of this world of many many, oh so many options available to us.
So since I was blatently ignorant on this, I thought I might share it with all of you.
Kind greats
Matthieu
|
|
Reply To Message
|
|
Author: SecondTry
Date: 2023-10-23 19:41
Ok, so I watched this video. I think I might be able to simplify it and its usefulness as applied to testing various clarinet setups.
You talk about your use of single variable design. This is the best way to test something that is changed by multiple variables. In simple terms, your changing just one thing, holding all other factors constant, leads you to best be able to assign whatever change you experience, to that thing you made variable. It practical speak, you might hold the reed, and testing music, and ligature constant while you change mouthpieces.
In the video's case the "reed" might be the "Potassium Nitrate," the "ligature" the "Sulfur," the "mouthpiece" the "Charcoal," and the "testing music" the Lampblack. You've held 3 of these 4 variables constant while you test different mouthpieces to see, if you'll pardon the video's metaphor, which one shines best.
|
|
Reply To Message
|
|
Author: Fuzzy
Date: 2023-10-23 20:08
I, too, watched the video, but am somewhat skeptical of its use as it applies to clarinet reed/mouthpiece/ligature combinations.
Here's why:
Place the mouthpiece on the clarinet, place the ligature, then place the reed. Tighten the ligature. Play.
Loosen the ligature, remove the ligature, reed, and mouthpiece.
Place the mouthpiece, ligature, and reed back on the clarinet. Tighten the ligature. Play.
Chances are great that the results will be different.
Rotation of the mouthpiece in relation to the clarinet body has most likely changed slightly, ligature placement on the mouthpiece has most likely changed slightly, and reed alignment on the table/facing has likely changed slightly. Likewise, torque/pressure on the reed has most likely varied. Also - perhaps the reed humidity has changed (if using natural cane).
I'm not certain how a person could obtain a reliable control set.
Fuzzy
;^)>>>
|
|
Reply To Message
|
|
Author: spikey1973
Date: 2023-10-23 22:08
@ Secondtry:
That the thing, I am no longer sure the one variable change is "the best" generally speaking, I our case, more likely then not though.
As Said, I am not sure myself, as this all is very complex.. just thought to share the idea about the principle.
@ Fuzzy:
You are right!
The obtainablilty of a reliable controle group is essencial here (in any research really). Although, if serious about testing, most of the things you mentioned can be remidied one way or another but how accurate and how accurate would be necessairy.. that is another mayor issue.
so oke,.. not usefull..
thanks for the responses though.
|
|
Reply To Message
|
|
Author: Tony F
Date: 2023-10-23 23:41
Apart from the variables that you are changeing intentionally, I don't see how you could avoid introducing unintended variables into the mix. Not rigorous enough for accurate objective testing.
Tony F.
|
|
Reply To Message
|
|
Author: kurth83
Date: 2023-10-25 23:55
That sounds overly pessimistic to me. It is easy to measure repeatability by simply repeating the measurement, and more specifically observing the deviation over multiple measurements. No meaningful study uses only one data point for each variation. All systems have variability, and measuring that variability is part of experiment design.
I suspect an added benefit of multivariate experiments is there may be unexpected synergies between the variables, I didn't watch the video and maybe they covered that. Personally I would do both single and multivariate measurements for that reason.
Aging classical trumpet player beginning to learn clarinet as a second.
|
|
Reply To Message
|
|
Author: Hank Lehrer
Date: 2023-10-26 03:27
Hi All,
Objective or subjective? Null or research hypothesis? Control group versus experimental group? Parametric or non-parametric? And the list goes on.
I do not see where the OP indicated some sort of measurement scheme.
If this stuff was easy, everyone would be doing it. One must ponder excessively about how to proceed before stepping off the curb.
Hank
|
|
Reply To Message
|
|
Author: kurth83
Date: 2023-10-26 05:22
In a way we are all always doing it, everytime we make a judgement that one reed or mouthpiece is better for us than another.
Most of the above is just putting fancy vocabulary on what we all do, search for what works for us. :-)
I confess I have done it systematically sometimes, and not so systematically other times. Results vary, but I usually ended up better off than I was before.
In some cases, over time I did get an instrument (family) dialed in to where I had exactly what I wanted. It took time and money because I had to develop a good understanding of what the tradeoffs were and where I wanted to sit on the various tradeoff curves. It was usually a combination of research and experimentation.
It also took a lot of practice since I discovered that first impressions aren't always accurate, even iof they are repeatable in the short term. I find they are right most of the time, but wrong often enough I like to sit with a change (or a set of changes) for a while to see how it grows on me.
Aging classical trumpet player beginning to learn clarinet as a second.
|
|
Reply To Message
|
|
Author: Hunter_100
Date: 2023-10-27 00:04
How do you quantify the results of this analysis? In the video, a numerical value is placed on the result of each test and these numbers are added up to determine the proportional improvement from each variable.
I struggle to come up with a good way of assigning a numerical value to the clarinet playing experience. Sure, you could say 1-10 on responsiveness, tone, articulation etc. but the numbers would likely be too subjective, and also prone to predetermined bias to yield a useful result. Perhaps intonation would be useful to study with this multivariate approach since we can easily determine pitch and put a number to this.
|
|
Reply To Message
|
|
Author: Mojo
Date: 2023-10-27 17:38
Fuzzy:” Place the mouthpiece, ligature, and reed back on the clarinet. Tighten the ligature. Play.
Chances are great that the results will be different.”
As kurth83 replied, evaluating repeatability (or noise) is usually part of experimental design. There is also something called “hidden replication” in a 2-factorial design. The analysis comes up with a noise number that your results have to overcome in order to be significant.
As Hunter 100 said, coming up with a number to quantify your results is also a challenge. You can try to rank your results on a. 1-10 scale of opinion. But these test designs work better on studies where you have an unbiased meter to quantify the results.
I did several of these kinds of studies as an engineer working on glass forming processes. But I have not found them useful for understanding what I like in a woodwind set-up. But I can not say I have tried to apply them either.
MojoMP.com
Mojo Mouthpiece Work LLC
MojoMouthpieceWork@yahoo.com
Post Edited (2023-10-27 17:39)
|
|
Reply To Message
|
|
Author: Hank Lehrer
Date: 2023-10-27 20:43
Hi All,
IMHO what we have here is an attempt to use nominal data for a quantitative statical analysis. The OP wants to do inferential statistics while the data are not useable. From the article cited below "Nominal data is labelled into mutually exclusive categories within a variable. These categories cannot be ordered in a meaningful way."
Here is a quick review of the properties of various numbers for those that may have forgotten the first few lessons of Statistics 101.
https://www.scribbr.com/statistics/nominal-data/
Hank
PS Remember that data is plural and datum is singular.
Post Edited (2023-10-27 23:18)
|
|
Reply To Message
|
|
Author: Fuzzy
Date: 2023-10-27 22:35
Hi Mojo,
Oh, I have no doubt such tests are possible. (There's no lack of evidence that tinkerers will tinker when given the time and/or money to do so.)
What I doubt is the general usefulness of such an endeavor...especially when coupled with the human body.
Fuzzy
;^)>>>
|
|
Reply To Message
|
|
Author: SunnyDaze
Date: 2023-10-27 23:27
Hi Matthieu,
I did a whole load of this kind of stuff at the beginning, but after I got my clarinet properly serviced by a good tech, and got a really good mp, I didn't need to do it any more. Unfortunately, I think that getting properly set up is kind of expensive, but after that, a lot of things just seem to work.
Sorry, I'm not sure that that is super helpful if you are on a tight budget.
Jennifer
|
|
Reply To Message
|
|
Author: JoeRomano
Date: 2023-11-06 03:29
This is a very plausible approach to doing some comparisons.
Anyone who is pessimistic would have to agree that when anyone is testing say just one component, like a change in mouthpiece, then one acquires information that is useful, even if not perfect. After all, we do rate in some partially objective and subjective way how it feels, responses, sounds and then make a decision about which to use. Same with reeds (brand, type, strength), ligatures, clarinets, etc. Designing an orthogonal array type of experiment not only
is an efficient may to reduce the total number of comparisons required, but
it can also quantify if any differences are real. That is not to say that more data wouldn't be better. It always is best to have more data. But we can learn from data that might appear to be insufficient.
|
|
Reply To Message
|
|
Author: alanporter
Date: 2023-11-06 04:11
Looking at your original message on October 23, I am wondering about line 5. What have your testes got to do with clarinet affairs ?
tiaroa@shaw.ca
|
|
Reply To Message
|
|
The Clarinet Pages
|
 |
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
