Hacker News Reader

Comment by todd8 | Hacker News Reader

The headline exaggerates the gains for women. From the article:

    This year: total. = 135992 up 31% over last year
               female =  38195 up 39% over last year

From this we can calculate last years numbers:

    Last year: total  = 103811
               female =  27478

Comparing total participation:

    This year: female/total = 26.5%
    Last year: female/total = 28.1%

African-Americans and Hispanic and Latino students made more robust gains. The gains for women might not even be statistically significant.

View on HN

That's a roughly 10% increase over last year in the relative rate of participation:

   (38195-27478)/(103811)*100 = 10.32%

Comparing two percentages directly like that, while intuitive, misses trends in relatively small subsets of the larger population.

It would be useful to calculate the percentage difference between n_group/n_population and n_group_course/n_course. This would essentially calculate a "representivity" score for that course.

So, for example, according to the 2010 US Census, Black or African Americans constitute 12.6% of the population. However, according to [1], they constitute only ~5% of the CS degrees awarded. Therefore the score for CS and Black or African Americans is |12.6 - 5|/12.6 = ~60% underrepresentation.

[1] https://datausa.io/profile/cip/110701/#demographics

This negates the idea that a certain demographic may be disinclined to a particular subject (which is equivalent to being inclined towards other subjects).

One needs access to courses without unnecessary discrimination, one does not need to be pushed in to courses simply to make their representation scores better.

Reply on news.ycombinator.com