|
|
This topic comprises 4 pages: 1 2 3 4
|
|
Author
|
Topic: Film Done Wrong - VERY WRONG
|
|
|
|
|
David Stambaugh
Film God
Posts: 4021
From: Eugene, Oregon
Registered: Jan 2002
|
posted 02-27-2003 10:57 AM
Maybe I'm missing out on some subtle distinction between mergers and acquisitions, or maybe I'm just splitting hairs. If two viable companies like Compaq and HP decide to join forces, that's clearly a merger. If one viable company (Regal) buys the assets of a non-viable company like Edwards, isn't that clearly an acquisition? Or if Microsoft buys some small company like Visio, that's an acquisition, not a merger. Philip Anschutz (Regal Cinemas) acquired the assets of Edwards and UA, lumped them in with Regal, and re-christened the resulting entity Regal Entertainment Group. Smells like an acquisition to me, not a merger, but then I failed Business Economics 101. ![[Wink]](wink.gif)
Was UA in financial trouble like Edwards was?
| IP: Logged
|
|
|
|
|
|
|
|
Evans A Criswell
Phenomenal Film Handler
Posts: 1579
From: Huntsville, AL, USA
Registered: Mar 2000
|
posted 02-27-2003 04:39 PM
OK. I'm back from my meeting and now I can elaborate.
The number of visits (or samples) required to determine a mean value of some attribute of a theatre (like projection quality, screen brightness, etc.) is based on how confident you want to be of your measured average. Saying that, for example, you want to be 95 percent confident that your measure is correct. Then the confidence limits are the mean plus or minus (1.96 times the standard deviation divided by the number of visits).
So, when you take samples, you must have a way of consistently and fairly assigning a numerical value to each sample. You'll need to calculate the average and the standard deviation of the samples and you'll need to kow the number of samples you have.
Let's use a real life example. I've been to the Carmike Century 8 in Decatur 61 times and the average presentation quality score (using a straight arithmetic average like you use to average grades in school) is -6.95, with a standard deviation of 6.70 . (The same logic would apply if I'd made 61 visits to all different Carmike theatres to evaluate the chain somehow).
So the 95 percent confidence interval has maximum
-6.95 + 1.96 * 6.70 / sqrt (61)
and minimum
-6.95 - 1.96 * 6.70 / sqrt (61),
which gives us the range -5.27 to -8.63 .
Therefore, I can say with 95 percent confidence that the actual presentawtion quality score, based on my 61 visits, is between -5.27 and -8.63 .
If I want to claim that it's in a smaller interval than that, then I have to reduce my confidence. The point is, with just 61 visits (out of thousands), with 95 percent confidence, I can make this claim. If I drop my acceptable confidence level to, say, 80 percent, then my interval will be smaller (use 1.28 instead of 1.96 in the above calculation, and you'll get the smaller range -5.85 to 8.05 .
You get a "law of diminishing returns" by sampling more since the interval's width does not decrease linearly with the number of samples, but with the square root of the number of samples. That means that having 610 samples instead of 61 would reduce the interval's with to just under one third its current range, not by a factor of 10, so taking a sample of 60 percent of a large population would be drastic overkill!
Using the red and green balls in a large, large jar as an example (assume it has an infinite number of balls in it and you know they're all red or green), if you draw out just 10 balls and 3 are green and 7 are red, after drawing out these 10 balls, you can say with 80 percent confidence that the percentage of red balls is between 51 and 89 percent. In other words, you already know with a great chance of being right that well more than half the balls are red.
When visiting theatres, once I get 10 visits to a location in, I feel I have data that has enough mathematical backing to begin using.
| IP: Logged
|
|
|
|
|
|
|
|
Evans A Criswell
Phenomenal Film Handler
Posts: 1579
From: Huntsville, AL, USA
Registered: Mar 2000
|
posted 02-27-2003 10:55 PM
The samples need to be random. If you sample 61 theatres, they need to be randomly chosen from the entire population of the chain's theatres.
Also, this logic assumes that the population size is, for all practical purposes, infinite (like sampling 61 showings out of thousands). If the population is small in relation to the number of samples, then, of course, the confidence would be much higher (and this method of calculating the confidence would not apply).
If a chain had 61 theatres and you sampled 61 different theatres thoroughly enough to determine their "scientifically calculated "suckiness factor", then you haven't taken a sample, but a census!
Assuming the population you're sampling to measure is infinite, then if you sample a small number of items from the population, then the standard deviation controls the confidence interval. If you sample 61 items from an infinite set, then the 80 percent confidence interval is the average plus or minus (1.28 times the standard deviation) / sqrt (number of samples) .
Therefore, if I sample just 10 entities from an infinite set, and get an average of, say, 90.0, with a standard deviation of 1.0, then my interval is 90.0 plus or minus (1.28 * 1.0) / sqrt (10) , which gives the interval 89.6 to 90.4 . With this, I could say with 80 percent confidence that the true average is between 89.6 and 90.4 . If I said, based on these 10 measly samples, that the true average in this case were between 89.6 and 90.4, I'd only have a 1 in 5 chance of being wrong! However, let's say that the standard deviation were much larger, like 10.0. Then my range would have been 85.95 to 94.05 . In this case, my statement for 80 percent confidence would not be nearly as powerful. To get the same range as I had with the 1.0 standard deviation case, I'd have to have 100 times the samples as before, meaning I'd have to have 1000 samples to achieve that same interval as I did with 10 in the 1.0 st. dev. case!
The explanation is intuitive. If something is varying a lot from sample to sample, it's going to take many more samples to get a good idea of the average. If something is being very consistent over time, then it takes far fewer samples to be confident of the true behavior on the average. The population being infinite or very large does not enter into this discussion. A small population only makes the situation better.
| IP: Logged
|
|
|
|
Evans A Criswell
Phenomenal Film Handler
Posts: 1579
From: Huntsville, AL, USA
Registered: Mar 2000
|
posted 02-28-2003 09:56 AM
It's really funny that we got into a mathematical discussion of "quality of suckiness". :-)
One reason I started the mathematical discussion is when people develop an opinion that something "sucks", it has to be based on something. When asked why you believe something "sucks", I believe you should be able to back it up with concrete examples. For example, if I say that Carmike 8 in Decatur has more presentation quality problems than any other theatre in the area, I feel I should be able to provide data supporting that fact. Now, there are few people out there as eccentric or weird as me out there that would keep track of presentation quality problems and write software to collect and put them on a WWW site with statistics, but doing so gives my opinion a lot more punch, provided that I haven't faked data or been unfair (I haven't faked data and try to be as fair as I possibly can).
Many people develop an opinion too quickly and when asked about their opinion, can only cite 2 or 3 examples. If I went to Carmike 8 only three times and got an average of 7 points worth of problems per show, that's not very meaningful, but getting that result after 61 visits is about 4.5 times as strong a statement (The 4.5 is sqrt (61) / sqrt (3)).
To make strong backable statements about an average measurable quality of something, you must first have a fair, consistent, meaningful way of assigning a numerical value to the quality of any sample. Second, you must take enough samples so that a small enough interval for a fairly good confidence level is achieved. If you have problems with either of these, your results may not be valid.
| IP: Logged
|
|
Scott Norwood
Film God
Posts: 8146
From: Boston, MA. USA (1774.21 miles northeast of Dallas)
Registered: Jun 99
|
posted 02-28-2003 02:23 PM
Evans hints at an important point: is "suckiness" a binary attribute--i.e. either a theatre sucks or it doesn't--or is there more to it than that? For me, it's more subjective than that.
A theatre with a lousy building, high ticket prices, lousy management, lousy concessions, lousy seats, and lousy film presentation definitely "sucks." A theatre with an attractive, well-maintained building, reasonable ticket prices, attentive management, tasty and inexpensive concessions, comfy seats, and "film done right" clearly doesn't "suck." But what about theatres which have great film presentation and rude employees who have no idea what "customer service" actually means? Or those where the lobby is impeccibly clean but the prints are filthy?
Very few theatres excel in all of these areas, and some vary considerably from day to day based upon which employees happen to be there at any given moment. Even generally good theatres can have problems (which might not even be their fault, but which still inconvenience the customer), and even lousy theatres can put on good shows on opening weekends.
I don't have a good solution to this, but felt it was worth presenting another problem with the idea of measuring the "suckiness" (or lack thereof) of any large theatre chain.
| IP: Logged
|
|
Evans A Criswell
Phenomenal Film Handler
Posts: 1579
From: Huntsville, AL, USA
Registered: Mar 2000
|
posted 02-28-2003 02:57 PM
Good point, Scott. "Suckiness" is not a binary attribute, unless you set a threshold and say all presentations that don't suck have a score above some number x and all others do. (I love this terminology, using the word "suckiness" rather than "problem index" or something).
I like for my measurements to be based on both the type of problem and the seriousness of it (how distracting it is during the movie or how much it detracts from the moviegoing experience).
I have a set of categories and the points deducted are slight if a problem is minor, and more drastic if the problem is severe. Let's say that everything is OK about an image except the aperture shadow causes, say, the entire top edge to be faded or jagged. That's a 2 point deduction. If the fading caused was very slight, or only caused a small part of that edge to appear that way, it would be a 1 point deduction. If all 4 edges looked really bad, that would be 8 points. If 2 were bad and 2 less severe, that would be 6 points. Running a movie significantly out of frame, affecting composition, is usually 10 points or more. I've been doing this a long time and have a good set of categories and point-assignment techniques for problems based on distractions. A presentation with no defects that warrant deductions gets a perfect "0" score. The worst score I've ever assigned was a "-35". See Dogma, Carmike 8 in Decatur, under the presentation quality reviews section.
So, I've got a system that gives a much smoother measure than a binary "this presentation is OK" vs. "this presentation sucks".
I've never documented this system thoroughly in a way on my site that lets site viewers understand the system very well. I keep meaning to do it, and I can't remember if I've enumerated the categories and "problem codes" that are used. I beleive I have, so I won't do it again here.
The reason you want a more continuous system is that if you have a binary attribute, every value is a 0 or 1. If you flip a fair coin enough times and get a 0 for heads and 1 for tails each time, and average this "measure", you'll get something close to 0.5 in all likelihood, which isn't close to any actual value measured. The standard deviation will be very close to 0.5, though. The wide standard deviation should be a clue that there is a wide deviation in the values.
Suppose at a company there are 9 employees that make $10000 per year and one employee that makes $1000000 per year. I can truthfully say "The average salary at the company is $109000 per year." That average isn't very meaningful, though. The standard deviation of $93920.12 should raise a big flag. A sanity check for a bunch of data is to compare the median with the average. If there two aren't fairly close, then maybe the average isn't what needs to be used for analysis. Remember, the average doesn't have to correspond with the value that occurs most frequently.
Now let's go home to our average families with 2 and a half kids. ![[Smile]](smile.gif)
| IP: Logged
|
|
|
|
All times are Central (GMT -6:00)
|
This topic comprises 4 pages: 1 2 3 4
|
Powered by Infopop Corporation
UBB.classicTM
6.3.1.2
The Film-Tech Forums are designed for various members related to the cinema industry to express their opinions, viewpoints and testimonials on various products, services and events based upon speculation, personal knowledge and factual information through use, therefore all views represented here allow no liability upon the publishers of this web site and the owners of said views assume no liability for any ill will resulting from these postings. The posts made here are for educational as well as entertainment purposes and as such anyone viewing this portion of the website must accept these views as statements of the author of that opinion
and agrees to release the authors from any and all liability.
|
Home
![[Big Grin]](biggrin.gif)
But what if Acme Cinemas operates theatres nationwide, and has excellent managers and technicians in every area of the country EXCEPT the Pacific Northwest. You sample every Acme location in the Northwest and find many problems. You've now sampled a fair percentage of Acme theatres, but only in one geographical region. How can you extrapolate from that sample and draw conclusions about Acme as a whole?
Printer-friendly view of this topic