2008-06-08, 06:40 | Link #242 |
out of touch with anime
Join Date: Sep 2006
Location: Greece
Age: 44
|
I need a clarification about the TV ratings and how they are conducted, due to my lack of Japanese.
Special boxes are given to the households that are connected to the TV set to see exactly which channel the viewers are watching? If so, exactly what percent of the households own that device? 5 % or more? Or is the survey filled through questionnaire, phone or online? I'd interested to see the sample population of this research. |
2008-06-08, 07:51 | Link #243 |
AS Oji-kun
Join Date: Nov 2006
Age: 74
|
Ratings methodologies are discussed in some detail earlier in this thread; I have some postings on the subject as does rooboy. Unfortunately the ratings methodology document that I mentioned there has vanished from Video Research, Ltd.'s web site. I didn't see a replacement item, but I didn't look too hard. The site map for the English site is here; a map with sample sizes by locale can be found here.
Video Research uses "people meters" as does Nielsen in the US. Meters are attached to every television in the home and collect viewing data continuously throughout the day. Buttons on the device are pressed by different household members to indicate that they are in the room at that time. I hope you've taken a statistics course so that you understand what matters in sampling is not the fraction of the universe that is sampled (your 5% figure above), but the total size of the sample. Accuracy only improves in proportion to the square root of the sample size, so you need to quadruple the sample to double your accuracy. Since costs grow linearly there are real limits on how much more accuracy is worth. Most ratings services use a sample in the vicinity of 2,000 households.
__________________
Last edited by SeijiSensei; 2008-06-23 at 23:50. Reason: various diddly typos, etc. |
2008-06-08, 11:04 | Link #244 |
out of touch with anime
Join Date: Sep 2006
Location: Greece
Age: 44
|
Thanks for the detail.
I asked because I thought there would be also other factors available except age and gender, eg like education, profession , wages, number of children, married or unmarried couples etc. But it seems ithis is not available to the public. This would be usefull especially for the late night shows. Often you get the impression in Japan that shows with 1% TV rating are watched by the whole population. |
2008-06-12, 20:46 | Link #247 |
Senior Member
Join Date: Jan 2007
Age: 32
|
sorry if this is off topic
Is it true that even though Japan made anime... the people in Japan dont really have much interest in it? My friend tells me that people in Asia are more interested in it especially places like Hong Kong and Gong Zhou |
2008-06-12, 21:03 | Link #248 |
Gamilas Falls
Join Date: Feb 2008
Location: Republic of California
Age: 46
|
Might be a little like living near a major tourist attraction or theme park. You yourself may never go there, or go there once because "It is always there" or "Yeah, so its right over there. Nothing special about it".
People who live in Orange Country have that kind of feeling about Disneyland, and people who live in the San Francisco Bay have that feeling about Pier 39.
__________________
|
2008-06-17, 22:49 | Link #249 | |
Senior Member
Join Date: Jan 2007
|
Quote:
Oh, and for those interested, here's some more ratings numbers from the past. These are the Top 10 Anime Rankings for the week of February 28th to March 6th, 1994: 01. Dragon Ball Z - 26.8% 02. Sazae-san - 26.1% 03. Crayon Shin-chan - 25.4% 04. Kiteretsu Daihyakka - 22.5% 05. Slam Dunk - 21.4% 06. Yu Yu Hakusho - 20.6% 07. Tsuyoshi Shikkari Shinasai - 19.9% 08. Doraemon - 17.8% 09. Tico of the Seven Seas - 16.0% 10. Sailor Moon R: The Last Episode - 14.4% Source: http://bbs.enjoykorea.jp/tbbs/read.p...73603&tab=five |
|
2008-06-22, 03:07 | Link #251 | |
out of touch with anime
Join Date: Sep 2006
Location: Greece
Age: 44
|
Quote:
Last edited by petran79; 2008-06-22 at 03:23. |
|
2008-06-22, 18:09 | Link #252 | |
Translator, Producer
Join Date: Nov 2003
Location: Tokyo, Japan
Age: 44
|
Quote:
But not having a new YPC5 episode was kinda sad... They made up for it, though, since this week's episode was pretty decently animated.
__________________
|
|
2008-06-22, 18:27 | Link #253 | |
Senior Member
Join Date: Jan 2007
|
Quote:
Not sure what the record is, though I thought I read somewhere that the first Astro Boy series in the 1960's scored a 40% rating one time. |
|
2008-06-23, 01:53 | Link #254 | |
Junior Member
Join Date: Jun 2008
|
Siegel Clyne, Destinyblade, SeijiSensei, and manny others...
I want to thank you for the reaserch you have givin and shared here. being new, I started with this thread after googling a favorit t.v. Japanese show and hoping it went up in the ratings. and instead saw it droped... can i be wrong? Quote:
droping out like that is not bad...right? to add a thank you to everyone who has put up much information. I am now subscribing to this thread for its informative content and would like to know what anime is on t.v. Last edited by Awazo; 2008-06-23 at 01:58. Reason: to add a thank you |
|
2008-06-23, 02:48 | Link #255 |
Lets be reality
Join Date: May 2007
|
It happens, it's late at night and MBS air it the day before so you can watch it on NicoNico or grab the raw and watch it Sat afternoon instead of staying up to 2 am or even staying home to watch. A lot of other shows dropped as well that week and the week before that....
*1.6% (*4.1%) 05/24 (Sat) *1:23am-*1:53am TV Tokyo Golgo 13 Also you really shouldn't pay too much attention to tv ratings for shows aired late at night, Code Geass and Haruhi were in the 1-2% range and those 2 shows are some of the best selling tv anime series on dvd period. Frontier is selling a lot of cd singles and albums and it's dvd/blueray preorders are pretty damn good too. |
2008-06-23, 04:55 | Link #256 | |
out of touch with anime
Join Date: Sep 2006
Location: Greece
Age: 44
|
Quote:
Problem is that back then in the 60s and 70s the number of televisions and phones was more limited and so many adults and kids visited friends or relatives to watch their favourite show. Which means that 40 % back then, could actually mean 70-80 % in today's standards. So Dog of Flanders, though with a slight margin, may be the first in popularity. Also Battleship Yamato had to be cancelled because Heidi had more TV ratings. So a similar situation existed back then too. Add also the limited number of TV channels so you had not many alternatives, even though the few you had were good. Even so, the less TV channels, the more time you have to do something better. I think the popular shows back then must have been around that rating. Many of them were innovative for their time (and some still are). Also in other countries where anime series were exported or co-produced, the shows reached similar ratings. (Eg shows like Heidi and Mazinger in Spain or the success of shows like Candy worldwide). Also had to do with the fact that there were only 2-3 TV channels and VHS was unknown. Whereas now with so many shows and channels, such ratings are impossible to reach. Not even events like Wold Cup finals or music contests reach such ratings. If anime today is more popular it is also more fragmented. Stll if you dont have any idea about anime, it is better at least to know it through the older series, rather than Pokemon. |
|
2008-06-23, 23:24 | Link #259 |
AS Oji-kun
Join Date: Nov 2006
Age: 74
|
This is a complicated question. I'm going to compress about half a semester of basic statistics into a single AS posting.
Let's imagine that we take a sample of 2,000 homes and find that 4% of those homes (or 80 in total) are watching a particular show. Now imagine we take another sample of 2,000 homes and find in that sample that 72 of the homes, or 3.6%, are tuned to the show. In another sample we find 84 viewers, or 4.2%. If we could carry out this type of "repeated sampling" and than plot the frequency distribution (or "histogram") of these estimates, we'd find they fit the "normal" distribution, that well-known bell-shaped curve. Statisticians measure sampling accuracy using the standard deviation of this distribution of successive estimates. A profoundly important result in statistics called the Central Limit Theorem actually guarantees that this distribution of estimates will exactly fit the normal curve. Knowing this fact, we can calculate the standard deviation for this distribution of viewing estimates in repeated samples. That quantity is formally called the "standard error of estimate" and is the basis for those "margin of error" values that are reported for public opinion polls. The standard error depends on two quantities. It's proportional to the standard deviation in the population of the characteristic we seek to measure (viewing show X in this case) and inversely proportional to the square root of the sample size. The formula is Standard Error = Population Standard Deviation / sqrt(Sample Size) Now it turns out that, for proportions, the Population Standard Deviation depends only on the percentage watching. Let p be the proportion of people watching a particular show and N be the sample size. Then it turns out the standard error is SE(p) = sqrt[p(1-p)/N] So for your example of 4% (or 0.04 expressed as a proportion), we have the following calculation: SE(0.04) = sqrt(0.04*0.96/2000) = 0.0044 So if we took repeated samples of size 2,000 from a population where 4% are watching a show, they would describe a normal curve. One characteristic of the normal distribution is that 95% of the results willl fall within two standard errors of the mean. In our case that means we'd get numbers between approximately 0.04-2*0.0044 and 0.04+2*0.0044, or between 3.1% and 4.9%. So it's possible that, in the population, as few as 3% of the homes, or as many as 5%, are watching this show, but it's highly unlikely that the true viewing figure is as low as 2% or as high as 6%. Spoiler for Advanced material:
Can we reduce this range of error? Yes, by increasing the sample size, since the population standard deviation is determined entirely by p. The problem is that the standard error depends inversely on the square root of the sample size. So if I wanted to halve the error range, I'd need to quadruple the sample. Unfortunately costs don't have the same relationship to sample size; in social surveys, costs are generally linearly proportional to the sample size. So, in practice, you decide how big an error you're willing to tolerate and invest in sample size accordingly. 2,000 actually looks like a pretty good choice for programs that draw 4% or less of the audience. Even at 4% the 95% confidence interval is under 2, so a 4 is probably not really either a 3 or a 5 in the population, and definitely not a 2 or a 6. Finally, notice that the numerator of the standard error gets smaller and smaller as the population proportion heads toward zero or one. The maximum error is at 50%. There is no minimum error, of course; the confidence interval just continues to shrink as you get closer to the endpoints. I hope this made some sense. None of this stuff has much to do with intuition. You need to see and understand the math to appreciate the power of statistical sampling. You might also note that nowhere did I say anything about how large the population is. All these results assume an infinitely-sized population. Spoiler for More advanced material:
__________________
Last edited by SeijiSensei; 2008-06-24 at 01:56. |
2008-06-24, 00:44 | Link #260 |
Senior Member
Join Date: Jan 2006
|
Oh, I've made a mistake in my earlier calculation, I meant 1% for 2000.
Kinda forgot where exactly should 2 go in that formula (Population Standard Deviation / sqrt(Sample Size))*100%. PSD is considered to be 1/2 for rough error estimations. And don't forget that humans have extremely uneven distribution of tastes, complicating matters greately. (just try and measure anime popularity near Akihabara!). |
Thread Tools | |
|
|