These graphs do not indicate the raw count of each tag over time, but rather how its "Odds ratio" has changed.
The "Odds Ratio" is a measure of how much influence the tag has on the outcome. A value above 1 means it has a positive influence,
a value smaller than 1 means it has a negative influence, and a value of exactly 1 means that it has no effect at all.
For example, if under the graph "Revealing" you see that "Male" has a "Odds" of 1, being male has no effect on whether or not the character is "Revealing".
If you on the other hand see a value above 1, it means a character being "Male" means its more likely to also be "Revealing".
An easy way to think of it is, if a value is far from 1 (smaller than maybe 0.5 or higher than 2) it might be an indication of bias.
I will at an later date add the confidence interval for these odds as well.
Note that the odds above 100 are set to 100, so as to make the graphs slightly more readable, and usually odds that high have a very low confidence interval and should be taken with a grain of salt.
Note that these graphs are on a log scale, and that since all tags are not in all sets there are gaps in the data, sometimes resulting in no lines and only points being on the graph.
The "Odds Ratio" is a measure of how much influence the tag has on the outcome. A value above 1 means it has a positive influence,
a value smaller than 1 means it has a negative influence, and a value of exactly 1 means that it has no effect at all.
For example, if under the graph "Revealing" you see that "Male" has a "Odds" of 1, being male has no effect on whether or not the character is "Revealing".
If you on the other hand see a value above 1, it means a character being "Male" means its more likely to also be "Revealing".
An easy way to think of it is, if a value is far from 1 (smaller than maybe 0.5 or higher than 2) it might be an indication of bias.
I will at an later date add the confidence interval for these odds as well.
Note that the odds above 100 are set to 100, so as to make the graphs slightly more readable, and usually odds that high have a very low confidence interval and should be taken with a grain of salt.
Note that these graphs are on a log scale, and that since all tags are not in all sets there are gaps in the data, sometimes resulting in no lines and only points being on the graph.