The following (below the dashed line) is the population statistics from the last Mathesar Report I ever did. It was dated 06/15/2006 and the entire report can still be found on the Los Angeles board.
It is almost 3 years old and is for Los Angeles so one should be VERY careful about drawing conclusions regarding the present market in Las Vegas from this data. As BoneBoyBob remarked on this board, I definitely need to be able to obtain a new baseline before publishing a new report.
Almost nothing in an individual provider’s statistics depends on the population statistics except, of course, the expected price. I admit the expected price is at best a very rough estimate in the absence of a new baseline. The fact that nothing else in the individual statistics depends on the population statistics is why I am taking the step of offering at this time to provide individual statistics for those providers who are interested.
If there is sufficient interest from the community I may move ahead with updating my software and gathering the data for a full report.
The equation I gave earlier relating expected price to appearance and performance is a slightly simplified version of the actual equation given below, but it gives essentially identical results. I stumbled across it by accident. Note that the correlation coefficient between the logarithm of the actual price and the logarithm of the overall score is 0.77. This means that 77% of the variance in (the log of) actual prices is explained by the equation. No, this isn’t 100%, but I seriously doubt that it is possible to do better than this. At least, I haven’t been able to do better. Considering that providers set their own prices using very different criteria, I am amazed that those prices end up with a 77% correlation to the numerical appearance and performance averages from the reviews.
Note that I am in no position to ask TER to change their review practices. If they did ask, my first request would be for two performance numbers. One that goes from 7 to 10 depending on services offered and one going from 1 to 10 based on how happy the reviewer was with the service received, whatever that service might have been. Combining what the escort does with how well she does it in a single number is my greatest single peeve with the present system.
Note also that my analysis is an analysis of the reviews and not of the escorts. The distinction is important. I can do a good analysis of the review numbers, but I have no control over whether or not these numbers are a good representation of what any potential client will experience with any particular escort. It is TER’s job to ensure that the reviews are honest and accurately reflect an actual session. Even if reviews are completely honest, reviewers select escorts to see and escorts screen potential clients. This introduces distortions that can’t be removed.
-----------------------------------------
7 MARKET REPORT FOR INCALL COMPANIONS
-------------------------------------
The following is obtained by analyzing the data for the 155 companions who have
a 60 minute incall escort price, 60 minute incall anal price, or 60 minute
incall massage and blow job price in their profile and whose primary city is
either Los Angeles or Orange County.
The following shows the distribution of prices for a 60 minute incall session.
$101 thru $125: 01
$126 thru $158: 04
$159 thru $199: 10
$200 thru $251: 36
$252 thru $316: 35
$317 thru $398: 19
$399 thru $501: 40
$502 thru $630: 04
$631 thru $794: 04
$795 thru $1000: 00
$1001 thru $1258: 00
$1259 thru $1584: 02
The 1st quartile price for a session is $250.00.
The 2nd quartile (median) price for a session is $300.00.
The 3rd quartile price for a session is $400.00.
The mean price for a session is $342.45 (range is $125 thru $1500).
The mean appearance score is 7.58 (range is 3.68 thru 9.50).
The mean performance score is 8.00 (range is 5.66 thru 9.71).
The correlation coefficient between (ln of) price and (ln of) product, where
product is (a_avg * p_avg), is 0.77.
The regression equation for price as a function of appearance points and
performance points is:
price := ((a_avg * p_avg) ^ a) * b
where:
price is the expected price for a 60 minute incall session
a_avg is the average appearance score for last twelve months
p_avg is the average performance score for last twelve months
a is the first regression coefficient (= 1.550343 this report)
b is the second regression coefficient (= 0.5510278 this report)
and
^ is the power (exponentiation) operator.
The probability that the first coefficient is actually zero (i.e., that price is
not a function of the product of a_avg and p_avg) is 0.000 (less than 0.0005).
For this group (the equation should not be extrapolated to other groups) the
expected price for a 60 minute incall session based on a_avg and p_avg AND NO
OTHER FACTORS would be:
For a (10.0,10.0) the expected price is $694.80
For a ( 9.5, 9.5) the expected price is $592.64
For a ( 9.0, 9.0) the expected price is $501.16
For a ( 8.5, 8.5) the expected price is $419.77
For a ( 8.0, 8.0) the expected price is $347.83
For a ( 7.5, 7.5) the expected price is $284.75
For a ( 7.0, 7.0) the expected price is $229.91
For a ( 6.5, 6.5) the expected price is $182.71
For a ( 6.0, 6.0) the expected price is $142.55
For a ( 5.5, 5.5) the expected price is $108.84
For a ( 5.0, 5.0) the expected price is $081.00
NOTE: The term "this group" means the population for which a expected price is
calculated in this report. These statistics do not apply to outcall companions,
massage providers, or S&M providers. They do not apply to incall companions with
TER lifetime averages below 5.00. They do not apply to markets other than Los
Angeles.
-----------------------------------------
My head hurts!
I guess my appreciation of the girls in the hobby is similar to my appreciation of art. I can't explain it, but I know what I like! Lol!
But I look forward to you reports being published, Mathesar! It'll be interesting to see how well your mathematical analyses match my own intuitive opinions!
T.L.
Dear Mathesar,
1) the population in TER is not a probability sample, so there is no way in my view to have reliable statistics overall;
2) we cannot run statistical test to such population (Kolgomorov-Smirnoff);
3) information are based on self-selection by reviewer, so there is a huge element of distortion here;
4) The correlation coefficient between (ln of) price and (ln of) product, where
product is (a_avg * p_avg), is 0.77 = that 77% of the variance is higly suspected...high correlation is always result of too much collinearity in the variable, thus the variance is redundant...;
5) the equation in the regression modelling is also suspect of multicollinearity: the two regression coefficients carry the same information...in the multiple regression model of the form P = f(p avg1, p avg2....p avgn) ^ a) * b
where price is a function of the average of appearance, the average of performance and the two regression coefficients, there is no error term...so the model is incomplete by the absence of structural error parameter...also there is no controlling variables to isolate if the price is a real function of the two average...such as influence of the local market prices, the experience of a provider, the biases of the reviewer, etc...
In brief, with no control variables there is no way to establish if the regression equation measure what is supposed to measure...
6) so the expected price is just an approximation based on higly biased information...
7) a new baseline price won't solve the problem in my view as long the TER rating system is so biased...eliminating the self-selection problem in the data is in econometric a huge issue...the population statistics in the TER system is so distorted that as long we rely on appearance and performance there is not way to get reliable statistics...
8) 12 months is an arbitrary choice: data on appearance and performance need to be collected for the period a provided has been in TER; if you rely on only the last 12 months you introduce another element of distortion; a provided need to be evaluate on long term; some provider are great on short term; but once you run simple descriptive statistics on the entire TER period things change remarkably: also you can calculate the frequency distribution of review data to see how many reviews a provider get in a month; if you run this on K-Girl you will get a very high frequency (which is a function of their low price); if you run on other you get different result (again their price make the difference here)...so long term is the best way to control for all these difference...on average in the long term distortion in the date are lowered...
That said, you make a miracle with such biased data...
BM.
One is that TER changes review scores.
Two is that reviews are subjective and come from all ends of the range already.
Three is that I already find that guys aren't looking at the reviews and reviewers often enough and this would be another reason NOT to do the proper homework.
Four, I don't like the numbers in the first place and really have no desire to become a "statistic" in any further sense.
But, I do love me some Mathesar
xo, E
