Ratings, again...
-
As long as you're working on CP.Net... What if the article ratings showed the standard or average deviation along with the average? These values are pretty easy to calculate and can give you a good idea as to the accuracy of the beloved 'average' value. If you're worried about scaring people away with terms like "standard deviation", you could calculate the StdDev then scale to a 1-10 range and call it something like "confidence" and represent it with a little graph or something. Here's a little example:
Article A: 3,4,4,5,4,4,3,3,3,3,3,2,2,5,4,3,3
Article B: 5,5,5,5,5,5,5,5,4,4,4,1,1,1,1,1,1Take these two sets of hypothetical ratings, for example. They have the same average (3.41), but B should probably be considered a more highly-thought-of article. Those 1's in B look out of line with the the rest of the values - it should be rated much higher. The ratings for article A, though, make sense - it's mostly 3's and 4's. A avg: 3.41 B avg: 3.41 Through the magic of Excel, we get the Average and Standard Deviations: A AvgDev: 0.719 B AvgDev: 1.418 A StdDev: 0.870 B StdDev: 1.660 Each of these values is much higher for Article B. This confirms that the ratings for A are more consistent than those for B. You could interpret this to mean that A's average more accurately reflects the majority of its ratings than B's average does for B's ratings. This is all basic statistics, nothing crazy. And here's the "confidence" thing i was talking about. I'll just scale the StdDev to a 0.0 to 1.0 range. A Confidence: 0.692 B Confidence: 0.412 Or, A's confidence is at 70% while B's is at 41%.
Conf = (2.83-StdDev)/2.83. 2.83 is the maximum StdDev for data in the range 1 to 5 (i think :) ). (this "confidence" is just something i made up, it's not related to any real statistical calculation) it's friday, i'm bored -c
"What the hell are you getting so upset about? I thought you didn't believe in God". "I don't," she sobbed, bursting violently into tears, "but the God I don't believe in is a good God, a just God, a merciful God. He's not the mean and stupid God you make Him out to be". -- Joseph Heller Smaller Animals Software, Inc.
-
As long as you're working on CP.Net... What if the article ratings showed the standard or average deviation along with the average? These values are pretty easy to calculate and can give you a good idea as to the accuracy of the beloved 'average' value. If you're worried about scaring people away with terms like "standard deviation", you could calculate the StdDev then scale to a 1-10 range and call it something like "confidence" and represent it with a little graph or something. Here's a little example:
Article A: 3,4,4,5,4,4,3,3,3,3,3,2,2,5,4,3,3
Article B: 5,5,5,5,5,5,5,5,4,4,4,1,1,1,1,1,1Take these two sets of hypothetical ratings, for example. They have the same average (3.41), but B should probably be considered a more highly-thought-of article. Those 1's in B look out of line with the the rest of the values - it should be rated much higher. The ratings for article A, though, make sense - it's mostly 3's and 4's. A avg: 3.41 B avg: 3.41 Through the magic of Excel, we get the Average and Standard Deviations: A AvgDev: 0.719 B AvgDev: 1.418 A StdDev: 0.870 B StdDev: 1.660 Each of these values is much higher for Article B. This confirms that the ratings for A are more consistent than those for B. You could interpret this to mean that A's average more accurately reflects the majority of its ratings than B's average does for B's ratings. This is all basic statistics, nothing crazy. And here's the "confidence" thing i was talking about. I'll just scale the StdDev to a 0.0 to 1.0 range. A Confidence: 0.692 B Confidence: 0.412 Or, A's confidence is at 70% while B's is at 41%.
Conf = (2.83-StdDev)/2.83. 2.83 is the maximum StdDev for data in the range 1 to 5 (i think :) ). (this "confidence" is just something i made up, it's not related to any real statistical calculation) it's friday, i'm bored -c
"What the hell are you getting so upset about? I thought you didn't believe in God". "I don't," she sobbed, bursting violently into tears, "but the God I don't believe in is a good God, a just God, a merciful God. He's not the mean and stupid God you make Him out to be". -- Joseph Heller Smaller Animals Software, Inc.
You could also confuse the hell out of everyone with the equation for calculating the standard deviation: SD = √Σ(x - avg.(x))²/n Chris Losinger wrote: this "confidence" is just something i made up, it's not related to any real statistical calculation There are such things as confidence intervals that allow you to represent your data. We can say that a random variable will fall within these two ranges of my dataset(SomeNumber, SomeOtherNumber) with XX% confidence. These are calculations that have existed for a long time. Nick Parker
-
You could also confuse the hell out of everyone with the equation for calculating the standard deviation: SD = √Σ(x - avg.(x))²/n Chris Losinger wrote: this "confidence" is just something i made up, it's not related to any real statistical calculation There are such things as confidence intervals that allow you to represent your data. We can say that a random variable will fall within these two ranges of my dataset(SomeNumber, SomeOtherNumber) with XX% confidence. These are calculations that have existed for a long time. Nick Parker
yeah, i knew "confidence" was a reserved work, so to speak, in statistics. i just didn't remember what it was for :) . -c
"What the hell are you getting so upset about? I thought you didn't believe in God". "I don't," she sobbed, bursting violently into tears, "but the God I don't believe in is a good God, a just God, a merciful God. He's not the mean and stupid God you make Him out to be". -- Joseph Heller Smaller Animals Software, Inc.
-
yeah, i knew "confidence" was a reserved work, so to speak, in statistics. i just didn't remember what it was for :) . -c
"What the hell are you getting so upset about? I thought you didn't believe in God". "I don't," she sobbed, bursting violently into tears, "but the God I don't believe in is a good God, a just God, a merciful God. He's not the mean and stupid God you make Him out to be". -- Joseph Heller Smaller Animals Software, Inc.
No Problem, I forgot to mention in the last post that it was a really good idea. Thanks Chris :-D Nick Parker
-
As long as you're working on CP.Net... What if the article ratings showed the standard or average deviation along with the average? These values are pretty easy to calculate and can give you a good idea as to the accuracy of the beloved 'average' value. If you're worried about scaring people away with terms like "standard deviation", you could calculate the StdDev then scale to a 1-10 range and call it something like "confidence" and represent it with a little graph or something. Here's a little example:
Article A: 3,4,4,5,4,4,3,3,3,3,3,2,2,5,4,3,3
Article B: 5,5,5,5,5,5,5,5,4,4,4,1,1,1,1,1,1Take these two sets of hypothetical ratings, for example. They have the same average (3.41), but B should probably be considered a more highly-thought-of article. Those 1's in B look out of line with the the rest of the values - it should be rated much higher. The ratings for article A, though, make sense - it's mostly 3's and 4's. A avg: 3.41 B avg: 3.41 Through the magic of Excel, we get the Average and Standard Deviations: A AvgDev: 0.719 B AvgDev: 1.418 A StdDev: 0.870 B StdDev: 1.660 Each of these values is much higher for Article B. This confirms that the ratings for A are more consistent than those for B. You could interpret this to mean that A's average more accurately reflects the majority of its ratings than B's average does for B's ratings. This is all basic statistics, nothing crazy. And here's the "confidence" thing i was talking about. I'll just scale the StdDev to a 0.0 to 1.0 range. A Confidence: 0.692 B Confidence: 0.412 Or, A's confidence is at 70% while B's is at 41%.
Conf = (2.83-StdDev)/2.83. 2.83 is the maximum StdDev for data in the range 1 to 5 (i think :) ). (this "confidence" is just something i made up, it's not related to any real statistical calculation) it's friday, i'm bored -c
"What the hell are you getting so upset about? I thought you didn't believe in God". "I don't," she sobbed, bursting violently into tears, "but the God I don't believe in is a good God, a just God, a merciful God. He's not the mean and stupid God you make Him out to be". -- Joseph Heller Smaller Animals Software, Inc.
This won't help if we get some idiot voting 10 1's. It's also going to add a lot of storage/processing time. But - definitely something to think about. cheers, Chris Maunder
-
This won't help if we get some idiot voting 10 1's. It's also going to add a lot of storage/processing time. But - definitely something to think about. cheers, Chris Maunder
Chris Instead of showing the average just show the total. Thus even a 1 is useful. If someone gives me a 1 my total still goes up. Don't show the number of votes or the average. The advantage with this is that now nobody can artificially bring an article down. A nice total-count system is a lot better than an abused-average system. And perhaps one more rule. Only members with at least 150 posts and 2 months of membership are allowed to vote. This helps to avoid the [sign up, vote, sign up again...] situations. Regards Nish
_
If I am awake and my eyes are closed, it does not necessarily mean that I am thinking of naked women._
-
Chris Instead of showing the average just show the total. Thus even a 1 is useful. If someone gives me a 1 my total still goes up. Don't show the number of votes or the average. The advantage with this is that now nobody can artificially bring an article down. A nice total-count system is a lot better than an abused-average system. And perhaps one more rule. Only members with at least 150 posts and 2 months of membership are allowed to vote. This helps to avoid the [sign up, vote, sign up again...] situations. Regards Nish
_
If I am awake and my eyes are closed, it does not necessarily mean that I am thinking of naked women._
Nish [BusterBoy] wrote: Instead of showing the average just show the total. Thus even a 1 is useful. Good idea! At first i thought that sounded awful, since poor articles with lots of people voting would still get higher ratings than some good niche articles... But for a given category, the rating would still give a good impression of which ones were more favored - while at the same time reducing the rational for malicious down-rating of articles, since cross-category comparison would be useless anyway. Nish [BusterBoy] wrote: Only members with at least 150 posts and 2 months of membership are allowed to vote. :omg: 150 posts sounds like a huge number. Ok, maybe not for you :rolleyes: . But still, isn't that maybe overkill?_**
Sometimes i only remember, The days when i was young
Nowadays no one remembers when they were young and stupid...**_
ADEMA, The Way You Like It
-
Nish [BusterBoy] wrote: Instead of showing the average just show the total. Thus even a 1 is useful. Good idea! At first i thought that sounded awful, since poor articles with lots of people voting would still get higher ratings than some good niche articles... But for a given category, the rating would still give a good impression of which ones were more favored - while at the same time reducing the rational for malicious down-rating of articles, since cross-category comparison would be useless anyway. Nish [BusterBoy] wrote: Only members with at least 150 posts and 2 months of membership are allowed to vote. :omg: 150 posts sounds like a huge number. Ok, maybe not for you :rolleyes: . But still, isn't that maybe overkill?_**
Sometimes i only remember, The days when i was young
Nowadays no one remembers when they were young and stupid...**_
ADEMA, The Way You Like It
Shog9 wrote: since poor articles with lots of people voting would still get higher ratings than some good niche articles The point to be noted here is that poor articles wont get votes. Each vote counts. A 5 counts more than a 1, but a 1 still counts. Thus instead of 5 being excellent and 1 being poor we have 1 being good, 2 better, 3 still better 4 very good and 5 excellent. And a 0 [no-vote] means bad. If the total for an article is 0, it means that nobody found it useful. Nish
_
If I am awake and my eyes are closed, it does not necessarily mean that I am thinking of naked women._