Or if there is no variability for a user, say all C ratings, we can say fuck off
We’re all a bunch of @ass holes at heart, but I trust nobody is out to sabotage this little project.
If we wanted to track user responses, these forum polls are not the best method to do it.
There’s probably a google form or something similar where every answer is traced back to a user and would make the data tracking a lot smoother.
I envy your optimism of this place
Ok, so after an @ass load of digging I determined that neither standard deviation or variance are the correct methods of determining “Controversial” (or “Consensus” since that’s basically the inverse), but learning how to calculate consensus opinion is the best means to this end. Unfortunately there is no handy excel formula for determining consensus, like there is for standard deviation and variance. Fortunately though a number of scientific papers have been written on it and they determined the formula for consensus to be the below, which I’ve worked into my data and appears to be accurate (or at least as accurate as I can hope to get).
4 Likes
.
4 Likes
6 Likes
Dear lord a C? Really?
It’s pretty but kinda a boring composition imo
1 Like
Fussy Dutchman > Dear Lord
9 Likes
I can get behind this but DL a C?
It’s OK, but not my favorite.
Fussy is an easy A-minimum tier song.
5 Likes
Is this gonna take a long time?
Fuck I love jam bands
6 Likes
Only saying this in case you’re interested. Not going to push for it to be done here.
The type of data you’re collecting are ordinal, not continuous. That is, there is an order to the ratings but the space between ratings is meaningless. Compare that to continuous numbers where there is infinite space between then. Doing operations such as mean and standard deviation on this type of data isn’t best practice because the data aren’t fit for those operations.
The top reply in this post might explain it better.
D tier to me. The jam is the same 3-4 min jam 9 out of 10 times. It used to be A or B tier in 2011-2013
I agree with this. Each song is an ordinal data set that I have to ascribe integer values to so I can get a numerical result allowing me to compare the various ordinal data sets. Standard deviation equations are an improper method of comparison because they treat deviation from the mean as having a high level of deviation (ie 19 people voted “s” and 1 person votes “a” then the deviation is 9, but if it was 11 and 9 instead then the deviation is 1 even though we all would agree the first example shows a greater degree of consensus).
You also can’t just say “we’ll treat it like the higher the deviation the greater the consensus” because that doesn’t account for outliers at the ends of the spectrum. For example 10 “s” votes and 10 “f” votes has the same deviation as 10 “b” votes and 10 “c” votes even though obviously the second scenario shows greater consensus.
The formula I pulled is from a scientific paper titled “analyzing ordinal data with respect to extrema”
Here’s an example of that formula in action:
Data set 1 = 9 votes A, 1 vote B, 1 vote C, 9 votes D
Data set 2 = 1 vote A, 9 votes B, 9 votes C, 1 vote D
Both 20 votes with the majority at the extremes for sample 1 and the middle for sample 2. We would all agree sample 2 has higher consensus. The standard deviation on both is exactly the same at 4, but the consensus equation works out to .0737 for 1 and .6633 for 2, which is a significant increase and therefore higher degree of consensus
5 Likes
Hell yeah, taking care of business. Sorry if you said what algorithm you were using before and it didn’t SU&R and DMOR.
I didn’t mention the paper I pulled it from or anything like that, figured it was already too nerdy for a dbmb lol
1 Like