Wow. Just to be clear, the function above was derived only from the original four data points; I did not look up the July 31 result.
I re-performed the fit, this time using all five points, and the difference is minor:
[tex]r=108.827d^2-9352.11d+239225.8[/tex]
Here are the differences between the actual rankings and the computed rankings, using the revised function above:
July 26: +2.0
July 28: -12.1
July 29: +15.9
July 30: -5.8
July 31: -0.1
The fit of this parabola is so good that it makes me suspicious that the page rankings may actually be generated partly by mathematical extrapolation of previous trends, instead of purely as a measurement of the number of pageviews.
Using the revised best-fit function, the predicted rank on August 4 becomes:
[tex]r=108.827(35)^2-9352.11(35)+239225.8[/tex]
rank = 45215
which is a change of only three spots from the previous predicted rank of 45218.
- Doug
Kent-
"Holiday" factor included ?
How about allowing for those who are out fixing their discbine ?
YOS
TNAI
Actuarial Webbery Is Not Our Forte
08/01: 52,368
> 08/01: 52,368
Note that alexa reports that the rpls.com web traffic rating has pretty much found a plateau. If there were ever a case study in the effects of bad web design, this has got to be a prime example.
rpls.com
08/01/10 - 65,854
07/31/10 - 65,308
07/29/10 - 66,957
07/28/10 - 66,179
The data set so far
Just to keep it for the record, here is the data set so far for alexa's US web traffic rank for BeerLeg.com
08/01/10 - 52,368
07/31/10 - 53,893
07/30/10 - 56,601
07/29/10 - 59,554
07/28/10 - 62,675
07/26/10 - 69,640
Probably 08/04/10 would be a good day to recompute the curve fit to see what it predicts for the level at which the web traffic rank of beerleg.com will settle out and about when it will happen.
The data set so far
Hmm. That latest point is a bad outlier, compared to the rest. Is that just a reflection of the lighter weekend traffic?
- Doug
The data set so far
> Is that just a reflection of the lighter weekend traffic?
That could very well be.
Here's some page view data for the same days that might help you:
- 07/31: 8,441
- 07/30: 13,157
- 07/29: 14,135
- 07/28: 15,600
- 07/27: 14,651
- 07/26: 16,495
Note that the dates for the Alexa rankings that we have posted here are the dates the rankings were RETRIEVED, which actually are the results of the PREVIOUS day. The page view data above is for the actual days they occurred.
The data set so far
> Hmm. That latest point is a bad outlier, compared to the rest. Is that just a reflection of the lighter weekend traffic?
Good question. It will be interesting to see that the weekday traffic ranks look like.
What I get out of looking at the US traffic ranks that alexa generates for rpls.com is that there probably actually is some measurement being made daily. The semi-random oscillations don't exactly look like what some smooth predictive function would generate.
> I ran with that assumption, though, and instead of linear regression I modeled the data with a second-degree polynomial, which might be okay for short-term predictions.
>
> It turned out to be a surprisingly good fit to the four given points. Here's the equation of the best-fit parabola, assuming that the rankings are i.i.d.:
>
> [tex]r=108.886d^2-9355.40d+239271.1[/tex]
>
> where [tex]r[/tex] is the ranking and [tex]d[/tex] is the day of the month of July.
Here is a bit more of the series of web traffic rankings reported by alexa.com. The asterisk denotes a ranking based upon weekend traffic which is typically lower than that of a weekday at the same site.
beerleg.com
08/02/10 - 52,354 *
08/01/10 - 52,368 *
07/31/10 - 53,893
07/30/10 - 56,601
07/29/10 - 59,554
07/28/10 - 62,675
07/26/10 - 69,640
Just for comparison purposes, here are the rankings of the Administration Issues Discussion Forum for Geospecial Professionals.
rpls.com
08/02/10 - 66,261 *
08/01/10 - 65,854 *
07/31/10 - 65,308
07/29/10 - 66,957
07/28/10 - 66,179
Here is the more of the series of the US web traffic ranks reported by alexa. For curve fitting purposes, it looks as if the traffic ranks reported on Sundays and Mondays (for the weekend days preceding them, followed by an asterisk below) probably ought to be analyzed as a separate function.
beerleg.com
08/03/10 - 49,318
08/02/10 - 52,354 *
08/01/10 - 52,368 *
07/31/10 - 53,893
07/30/10 - 56,601
07/29/10 - 59,554
07/28/10 - 62,675
07/26/10 - 69,640
Note how much better the beerleg.com statistics are than those reported for another website.
rpls.com
08/03/10 - 68,088
08/02/10 - 66,261 *
08/01/10 - 65,854 *
07/31/10 - 65,308
07/29/10 - 66,957
07/28/10 - 66,179
Fitting a quadratic polynomial function to the beerleg.com data (less the weekend values) gives the following coefficients:
[tex]r = 295909.9 -13383.8d +180.24d^2[/tex]
where:
[tex]r[/tex] = alexa US web traffic rank
and
[tex]d[/tex] = days after June 31, i.e. August 1 = day 32.
Here's a graph of US web traffic rank over time:
The above function predicts that beerleg's US web traffic rank as reported by alexia will stabilize around this Friday at about 47,464.
and this means?
Will Amazon start advertising here?
> Will Amazon start advertising here?
Once search engines find your avatar, I'm thinking hair care products. :>
Everyone has a bad hair day here.:-)
but
If Wendell really wanted to "poke" the #'s, he would "link" the Beerleg store somehow to Amazon and start selling a variety of 'surveying sundries' besides BL T shirts, mugs etc...
maybe a "Kent the Surveyor" action figure
Estimation and curve fitting-Muddy
One of those dolls where you pull the string and it says "Best Regards" or such. 🙂
Not sure it would be a big seller in Lawton.
Estimation and curve fitting-Muddy
I think the new ones don't need a string.
You just shake it like hell and they talk. 😉
Action figures?
Okay, so a wookie and a cartoon figure of a homeless highway maintenance worker want some company in the fantasy action figure department? LOL!
Action figures?
HEY BACK OFF1
or
I am going to ask you to "put up your myths".
🙂
Action figures?
> HEY BACK OFF1
Hey, talk to the homeless guy picking up trash with a machete over there.
Action figures?
Hah...Your just worried if Wendell made you an avatar that it would draw the ire of the peanuts franchise over your obvious resemblance to Charlie Brown!
