Forum Home Forum Home > Site News, Newbies, Help and Improvements > Help us improve the site
  New Posts New Posts RSS Feed - Most polarizing albums
  FAQ FAQ  Forum Search   Events   Register Register  Login Login

Topic ClosedMost polarizing albums

 Post Reply Post Reply
Author
Message
Smurph View Drop Down
Forum Senior Member
Forum Senior Member
Avatar

Joined: January 11 2012
Location: Columbus&NYC
Status: Offline
Points: 3167
Direct Link To This Post Topic: Most polarizing albums
    Posted: September 05 2013 at 17:25
Could we set up a system that would also calculate standard deviation on all albums, therefore we would know the most polarizing albums automatically? It would be a cool option to look through in the search system.
Back to Top
irrelevant View Drop Down
Collaborator
Collaborator
Avatar
Honorary Collaborator

Joined: March 07 2010
Location: Australia
Status: Offline
Points: 13382
Direct Link To This Post Posted: September 06 2013 at 06:19
^ Would be! 
Back to Top
Dean View Drop Down
Special Collaborator
Special Collaborator
Avatar
Retired Admin and Amateur Layabout

Joined: May 13 2007
Location: Europe
Status: Offline
Points: 37575
Direct Link To This Post Posted: September 06 2013 at 08:35
I question how useful this would be (I like the idea in principle but...)
 
From a purely statistical standpoint the population sizes are too small and the ratings are too course to give any meaningful information in a single value (as are the averages in reality). Standard deviation refers to the variance from the mean in a normal symmetrical distribution (ie where mean, median and mode are equal) and ratings do not follow a normal distribution, they are always skewed or truncated, it is very (very) rare for an album to exhibit a normal distribution to the point of being essentially unheard of. For example ITCOTCK, CTTE, DSOTM, TAAB and SEBTP are all skewed or truncated distributions... and there is nothing wrong in this, it's what we would expect because people cannot give greater than 5 or less than 1 ratings to albums and they can only give whole-number ratings and because ratings are subjective and they are not actually comparative and our "population" is naturally biased and naturally polarised. So this means it is practically impossible for anyone to give a rating that is the mean, median and mode value. The standard deviation can be used to calculate the degree of skew in a distribution but you cannot do that from the mean and standard deviation alone, you need the median or the mode but even that is not enough to tell you that the ratings are polarised, or wrong, or buggered-up, or misleading.
 
The distribution graphs on each album page tell you a lot but they must be viewed with caution, they can reveal that opinion is polarised or divided but not why that is so, you cannot look at the graph and determine why the distribution is as it is. A standard deviation value tells you even less.
What?
Back to Top
Smurph View Drop Down
Forum Senior Member
Forum Senior Member
Avatar

Joined: January 11 2012
Location: Columbus&NYC
Status: Offline
Points: 3167
Direct Link To This Post Posted: September 06 2013 at 16:44
Originally posted by Dean Dean wrote:

I question how useful this would be (I like the idea in principle but...)
 
From a purely statistical standpoint the population sizes are too small and the ratings are too course to give any meaningful information in a single value (as are the averages in reality). Standard deviation refers to the variance from the mean in a normal symmetrical distribution (ie where mean, median and mode are equal) and ratings do not follow a normal distribution, they are always skewed or truncated, it is very (very) rare for an album to exhibit a normal distribution to the point of being essentially unheard of. For example ITCOTCK, CTTE, DSOTM, TAAB and SEBTP are all skewed or truncated distributions... and there is nothing wrong in this, it's what we would expect because people cannot give greater than 5 or less than 1 ratings to albums and they can only give whole-number ratings and because ratings are subjective and they are not actually comparative and our "population" is naturally biased and naturally polarised. So this means it is practically impossible for anyone to give a rating that is the mean, median and mode value. The standard deviation can be used to calculate the degree of skew in a distribution but you cannot do that from the mean and standard deviation alone, you need the median or the mode but even that is not enough to tell you that the ratings are polarised, or wrong, or buggered-up, or misleading.
 
The distribution graphs on each album page tell you a lot but they must be viewed with caution, they can reveal that opinion is polarised or divided but not why that is so, you cannot look at the graph and determine why the distribution is as it is. A standard deviation value tells you even less.

Understood. Dean you always give reasonable explanations- cant argue against this. Thought I would at least run the idea by.
Back to Top
 Post Reply Post Reply

Forum Jump Forum Permissions View Drop Down



This page was generated in 0.148 seconds.
Donate monthly and keep PA fast-loading and ad-free forever.