Wikipedia talk:Wikipedia Signpost/2017-12-18/Recent research


 * The plot of changes in the female ratio of bios and size. It's very interesting. A few points, though. First, is it supposed to be called: "Changes in female ratios and absolute numbers of biographies among the Wikipedias"? Is that what you mean? I first thought "size" meant average word/byte length of each bio. Second, great to convey three aspects in the one chart. But the steepness of the arrows, am I correct, is a red herring? The Norwegian- and Finnish-language WPs seem to start from a very low base, so I'm not sure how to interpret that. I don't know whether I care that the en.WP has a huge number of bios all-up, in this context. Still thinking about that.  Third, what are we to understand from the sharp decline in the ratio in the Chinese WP?  I'd love to see more gender-based graphs like this, exploring more detailed questions of gender. Tony  (talk)  07:27, 18 December 2017 (UTC)
 * I interpreted the slope of the arrows as showing the degree of concentration on adding female biographies. A high slope (jawiki) means that the added biographies concentrate more on females than the ratio on that wiki beforehand. A flat line would mean the added biographies are of the same ratio as existed beforehand. A downward slope (zhwiki) would mean the added biographies concentrated on men more than was already present on that wiki. Jujutacular (talk) 16:28, 18 December 2017 (UTC)
 * Yes, I think you're right. So we can match the slope with the rightward placement to get an idea of the composite impact, I guess. Tony (talk)  05:19, 19 December 2017 (UTC)


 * In the section "Assessing article quality and popularity across 44 Wikipedia language versions", I couldn't understand the graphs. Is it higher quality on the left or the right? What do the shades mean? I see the scales with their numbers but I don't know what they measure. SchreiberBike &#124; ⌨   02:41, 19 December 2017 (UTC)
 * Higher quality in those graphs are indicated by green/blue, whereas lower quality are red/orange. The "presumed" lower quality can be interpreted as near the bottom of the scale. Using the first plot on the left as an example, you're almost certainly going to encounter a stub article (in red) if the article length is around 337 bytes. At the top of the same plot, you're nearly always reading B-class or better articles. Same interpretation applies for other graphs. As the number of references in an article increases, the article quality tends to increase. We can see that number of images used in the article is not a strong indicator comparing to article length or number of references, since there are some GAs and FAs with as few as 2 and 3 images respectively. And this graph tells us that half of the articles have 5 images or less. Number of header sections tells us more headers does not always lead to better article quality since FAs tend to cluster at around 8 to 13 headers. More headers beyond that could lead to information overload or contents which should be spun off to a standalone page. And finally, the last graph on the right shows that more references used over a shorter article tends to have higher article quality, but only up to a certain point. OhanaUnitedTalk page 19:40, 25 December 2017 (UTC)