Incidentally, Earthling, what sort of fitting function are you using? And is it a theoretical model, or just an arbitrarily chosen functional form to fit the data? The largish deviation between the two curves for early points suggests that you’ve increased your degrees of freedom between the two fits. You do have significantly more points than degrees of freedom, right?
It’s not a theoretical model. I simply used Excel’s “add trendline” function for its charts to fit the data already available. I do wish that I had the brains and tools to do a better analysis. (Yo, kabbes! Check in here!) The first projection was plotted on a power curve using the formula that Excel supplied:
y = 826.48x[sup]0.246[/sup]
R[sup]2[/sup] = 0.9971
The new projection was plotted on a 5th-order polynomial curve with the formula (courtesy Excel, again):
y = 3E-08x[sup]5[/sup] - 2E-05x[sup]4[/sup] + 0.0056x[sup]3[/sup] - 0.7022x[sup]2[/sup] + 48.967x + 931.68
R[sup]2[/sup] = 0.999
I assume the “E” stands for powers of 10. If anyone wants to play with the data themselves, this is what I’ve been working with:
Time CMs (no SDSAB)
2004/03/23 02:10 AM 1002
2004/03/24 11:25 AM 1500
2004/03/27 06:26 AM 2005
2004/03/29 02:00 PM 2205
2004/03/30 09:30 PM 2300
2004/03/31 09:32 PM 2376
2004/04/01 12:00 PM 2430
2004/04/02 12:45 PM 2454
2004/04/03 12:31 AM 2492
2004/04/04 12:18 AM 2534
2004/04/05 12:02 AM 2575
2004/04/06 12:55 AM 2629
2004/04/07 12:05 AM 2669
2004/04/08 12:05 AM 2715
2004/04/09 12:38 AM 2750
2004/04/10 12:15 AM 2789
2004/04/11 12:05 AM 2822
2004/04/12 03:07 AM 2851
2004/04/13 12:35 AM 2900
2004/04/14 12:25 AM 2932
2004/04/15 12:15 AM 2968
2004/04/16 12:11 AM 3025
2004/04/17 03:30 AM 3081
2004/04/18 12:05 AM 3117
2004/04/19 02:55 AM 3167
2004/04/20 12:11 AM 3240
We just broke through 3,400, folks. 3,401 at 7:37 p.m. CDT, to be exact.
On the way to 3,500 with ~17 hours to go.
Thank you for that kind update. I wish we had one of those counters so we could watch the action.
Could you continue posting them until noon tomorrow? Possibly every hour on the hour?
Yes, seriously!
79 subscribers in the last 7 hours … not bad!
Might as well throw another number in, I’ve decided (after a month) to stay. I’ve thought about it, and thought of all the times that I wanted to post but held back.* Quite a few. Thought of the times I wanted to search on something, and that was a lot. Then I realized that I like it here and really wanted to stay. So:
NumberOfCharterMembers++; // **
- Yes, I know I could post as a Guest, but I wanted to see how many times I would post in an average month to see if it’s worth it.
** Assuming my credit card works with the subscription model here. If it doesn’t, the admins’ll hear from me!
Checking in for my (roughly) once-per-day data point, I see that we’re now at 3463 Charter Members (no SDSAB). We’ve added 223 CMs in the last 24 hours, so if this rate keeps up until the cutoff at noon, we’ll have 3500 licked easily.
Interesting. The power law worked remarkably well, fitting 20+ data points at a cost of only two degrees of freedom. Given that, I would stay well clear of polynomials (especially fifth order polynomials! Eek!), and try to extend the power law in some manner to include the upturn at the end. Perhaps a sum of power laws? The big problem, though, is that we have so few points in the upturn, so it’s hard to judge exactly what form the modification should take.
Another point, incidentally: The power law (or sum of power laws) is explicitly monotonically increasing and explicitly zero at the origin, while the quintic is neither. Since these are both features characteristic of the data, that should be reflected in our model.
Tomorrow afternoon, I’ll play around with the data some, see what I can come up with. In the meanwhile, it might be instructive to plot the results on a log-log scale (this will probably be under “axis options”, or somesuch), since a log-log scale will turn a power law into a straight line.
3484 charter members as of this post.
Make that 3485…
God licked my brain, ho-ha!
…um, this is the place for test posts, right?
Well Ed Zotti’s saying 3,503 over here which doesn’t tie up with my counting on the Show Groups page (I got 3489). So he has newer information than that page or I’m wrong or something.
In any case he’s delirious and we’ve passed the 35 hundred barrier.
SD
Ed has access to the actual data, including SDSAB and moderators that have paid. Indicating that 14 of them have paid.
I think its more likely that Ed’s count includes all 19 SDSABs and that in the time between Ed’s and SpaceDog’s counts 5 more members joined.
The point is now moot, as of 1000 CDT:
3,508 Charter Members
3,511 Charter Members + three confirmed paid SDSAB
3,527 Charter Members + all SDSAB
(Mods not included)
Your poor tongue.
It’s quite possible, but then one must wonder how you account for the SDSAB and moderators that we know have paid, due to them saying so.
Is it possible that some (poor fools) joined and got banned and therefore don’t show up in the Charter Member list but are still being counted by Ed ? Or that the Charter Member page is only updated every once in a while ?
Of course it’s entirely possible that I stuffed up the counting. I was using a wee linux script that I knocked together (while I should have been doing something far more profitable for my employer) to scan the HTML source. It’s not unlikely that I screwed up the count by a few members.
It all depends how it’s counted, in any case it’s still a staggering sign up rate, and still time before the new deadline.
SD
I don’t think the software distinguishes between paying and non-paying SDMBs. There are several categories of posters, such as Charter Member, Moderator or SDSAB.
To the software it doesn’t matter if you were moved to one of those categories automatically for paying the subscription fee or manually by an admin. Once you are a Charter Member or SDSAB you are treated the same as everyone else in that category.
I’m aware of two Charter Members who were banned, and their usernames don’t show up in the Forum Leaders view. If the software (or Ed) only looks at who have paid, however, then they would be counted into the subscriber population. But just these two won’t account for the large-ish discrepancy unless there have been more bannings that no one noticed.
Or, one or more of the other hypotheses so far posited are correct.