Rickresource Rickenbacker Forum

dpowell wrote:

FretlessOnly wrote: Well, I'm a scientist too (and I use statistics regularly), and I don't see how one can claim that the members aren't randomly "chosen." In fact, they are not chosen at all (which to me is irrelevant to this point); it is a voluntary thing that depends upon two factors:

1. interest in a particular brand; and
2. access to the internet.

Neither of these two population characteristics would seem to me to favor a particular finish, and further, neither would seem to favor a proclivity toward registering or not registering an instrument here. If anything, actually choosing a sample from the RIC population at large would be more likely to produce skew than would the voluntary joining of this website and registering one's instrument(s). To me, anyway.

I would agree that the members here who have registered their instruments may have a greater liklihood to own more rare finishes such as desert gold, azureglo, blue boy, teal, etc., but as to the big three, I think that our registry sample on these must be fairly representative of the population as a whole.

In a real scientific survey, they would be randomly chosen. I'm just saying that anyone who extrapolates production totals from the data without knowing the effects of those variables is at risk of error. YMMV.

That's not actually true. In fact, practically every scientific survey that has ever been done does not have a truly random sample. Yes, of course there's a risk of error -- but given the large number of instruments in the register, and no obvious reasons for large biases (i.e., the kind that would cause JG instruments to be only 50% as likely to appear in the Register as FG instruments), the chance that the proportions in the Registry are way off is likely to be extremely low. Nothing is absolute in the survey world, but you learn to be pretty comfortable with things that you're 95-99% (or more) sure of. In this case, I'd be pretty confident that the order of popularity of the finishes -- FG, JG, then MG -- is probably right, as well as the general proportions. I would not be remotely confident about the relative proportions of the rarer finishes because there's a good reason to believe the registry is biased towards including them.

Just because you don't know all of the variables doesn't make you can't make a very educated and likely rather accurate guess, especially if you've thought through possible sources of bias.

I own a Burgundyglo 4000 bass, but that's because that's all the store had available in 1978 when I bought it. I'm glad I have it, but it really wasn't a "choice." I just wanted a Rick!

EDIT: I suppose that doesn't really address the original question. Or does it?

I'd be pretty confident that the order of popularity of the finishes -- FG, JG, then MG -- is probably right, as well as the general proportions. I would not be remotely confident about the relative proportions of the rarer finishes because there's a good reason to believe the registry is biased towards including them.

I’m inclined to agree with this for the following reasons:

1) FireGlo was the default finish on pretty much every model from the date they coined the name. JetGlo accounts for instruments where the wood had some cosmetic issue preventing it from being sprayed FireGlo, so these were built without being custom ordered on a fairly regular basis. Have you ever seen a FireGlo (or any translucent finished) Rick with lousy maple? Nope. When I was an authorized dealer I sold many models finished in JetGlo that weren’t orderable that way, because of wood imperfections.(360/12V64s being an obvious example) MapleGlo was not a standard color for many years, it was a custom order. Most people wouldn’t wait months or years for a natural finished guitar when they could get a FireGlo one in days or weeks.

2) If it’s a rare color that’s turned up online anywhere in the past 15 years it’s very likely that one of this forum’s members either owns, owned, or held it in their hands at some point and likely jotted down the serial number somewhere, if not have it committed to memory. (I easily have the serial numbers for over 1,000 Rickenbackers in my files, with date ranges from the 1940s through 1996.) So a disproportionate number of custom color instruments will have made it to the register.

This is a fan site. The vast majority of the people who own Rickenbackers own one of their guitars or basses, and couldn’t tell you what names the company gave their finishes. Of course the results here will be skewed to some degree.

I was on the team of guys that helped put the registry together. Though a good portion of the instruments in the registry do belong to registered RRF members, a great number of them do not.

At this point i should assume that Mapleglo is the rarest color (between the standards of course).

To me it's interesting that MG would be the rarest standard color, since that is the finish I have seen mostly when I see a Rick bass for sale in my town. I have also seen JG basses, but don't think I ever see FG.

The main thing about any statistical analysis is, "How accurate do you want to be (accuracy), and how often do you want to be that accurate (level of uncertainty)?"

I did a run through with billiard balls as an example of random sampling of a small population of things. I will summarize this example below, but if anyone wants to see the entire statistical analysis, PM me and I'll send you my Excel spreadsheet.

Ignoring the cue ball, there are a total of 15 billiard balls consisting of two types, 8 solids (1-8) and 7 stripes (9-15). Thus there are 53.3% solids and 46.7% stripes in the population.

Now, let us put the 15 balls in a opaque bag and draw out differnt numbers of them as samples of the population. Let us first set a goal of wanting to be within 30% of the correct percentage in the population (accuracy).

Starting out with the lowest sample (n = 1), we either pull out 1 stripe or 1 solid, so our sample produces either 100% solids or 100% stripes; neither result is within 30% of the true percentage, so this sample is very inaccurate. Another slightly more statistical way of stating this result is that there is zero percent chance of obtaining a result with the desired accuracy with a sample of n = 1 (representing only 7% of the population).

Next, we sample two billiard balls at time (n = 2), resulting in either 2 solids, 1 solid and one stripe, or 2 stripes, so our sample produces either 100% solids (again), 50% solids/50% stripes, or 100% stripes (again). Oddly enough, the 50/50 sample is within our goal of 30% accuracy, but the other two are as above very inaccurate. Restated, with a sample of n = 2 (representing 13% of the population), we have a 1 in 3 (33%) chance of obtaining a result with the desired accuracy.

With n = 3 (representing 20% of the population), there is a 2 in 4 (50%) chance of obtaining the desired accuracy.

With n = 4 (representing 27% of the population), there is a 3 in 5 (60%) chance of obtaining the desired accuracy.

With n = 5 (representing 33% of the population), there is a 3 in 6 (50%) chance of obtaining the desired accuracy.

With n = 6 (representing 40% of the population), there is a 4 in 7 (57%) chance of obtaining the desired accuracy.

With n = 7 (representing 47% of the population), there is a 4 in 8 (50%) chance of obtaining the desired accuracy.

The statistics start to change a little with n > 7 simply because the number of solids and stripes limits what combinations are possible with a population of only 15 billiard balls (the sample represents more than 50% of the population).

With n = 8 (representing 53% of the population), there is a 5 in 8 (62.5%) chance of obtaining the desired accuracy.

With n = 9 (representing 60% of the population), there is a 5 in 7 (71%) chance of obtaining the desired accuracy.

With n = 10 (representing 67% of the population), there is a 6 in 6 (100%) chance of obtaining the desired accuracy; there is also a 100% chance for all samples with n > 10.

So, if you're still awake after being bored with numbers, what conclusion can we make? Mainly, this one: For samples numbering less than 50% of the population, we can assume that we have only about a 50% chance of obtaining a sample value that is within 30% of the actual population value.

I'll continue in the next post with my opinions of what all this means with regard to the Rickenbacker population.

Um, okay......

Actually I followed that but have no statistical background so could not have generated it. What I concluded? We'll never know.

O.K., so here's a cartoon representing the Rickenbacker population:
I've arbitrarily shown only three subdivisions (samples) of the population; the reason two are shown overlappig is that in reality they do overlap in the analysis that follows. Additionally, one makes the assumption (IMO, valid) that the population is heterogeneous with regard to the various models and finishes.

I did a statistical analysis of some recent eBay and GuitarBase listings. There were instances of the same instruments posted on both sites, so that's the reason for the overlap in the cartoon. Here is the breakdown:
(sorry, I can't figure out how to make this bigger on the screen.)

I've included the numbers Peter cited earlier in the thread for comparison.

So although the samples differ in n, the GuitarBase and eBay percentages are reasonably close. One could conclude that, within reason, these represent the distribution in the part of the population consisting of Rickenbackers for sale.

Now, looking at the Register values, it is interesting that the percentage of Jetglo is in the same ballpark as the others, and the Fireglo value isn't all that different, either. The big difference is in the percentage of Mapleglo, nealy a 10% difference. There are at least two possibilities: there are more MG instruments for sale by dealers (in other words, Mapleglo is over-represented in this sample), or the players/collectors of the Forum have fewer examples of MG in their arsenal of instruments (in other words, Mapleglo is under-represented in this sample).

In the long run, all such statistical analyses are just samples of a large population and represent "snapshots" of the population as a whole. It is up to the individual to ascribe validity to the numbers.

Nice work on the analysis, Gary!

AAAAAAAAgggggggguuuuhhhhhhh!!! That brings back nightmares of sadistics... uh, statistics classes in college!!!
Nice work though... Thanks for putting in the effort! I'm just glad I didn't have to do it...

Good job, Gary!

I like it! Nice work Gary!

Gary: Thanks for your additional work and the model you have provided with regard to the Rickenbacker population base.

I would add that the RickResource Register also has information from gbase and ebay and as such I would have its circle overlapping with each of these sources as well.

By way of additional comment the RickResource Register has data spanning many decades rather than the current year data that you obtained from gbase and ebay. That, in my opinion, is the reason for the discrepancy in some areas, in this case the percentage of mapleglo models.