3 ms·
Using a control group here doesn't make sense because we're not trying to benchmark SV against other places, nor are we doing an experiment. This is a survey.
by smt88 11y ago
Using a control group here doesn't make sense because we're not trying to benchmark SV against other places, nor are we doing an experiment. This is a survey.
Furthermore, if there are 400,000 tech workers in Silicon Valley, then 200 gives us a less than 8% margin of error (95% confidence level). This still gives us a result of at least 50%.
But we're not surveying the entire tech industry, we're surveying women in Silicon Valley, which is a number much closer to 100,000 (although that doesn't radically change our margin of error).
- brazzledazzle 11y agoHonest question: do those confidence levels account for how the participants are chosen? Or is it assumed that it's random?
- smt88 11y agoRandomness is assumed for my number. However, there are lots of ways this survey could have been biased: - women who were assaulted, harassed, or threatened have already left SV - women who were assaulted, harassed, or threatened were more likely to respond - women who don't believe in an anti-female culture in SV were more likely to respond - etc. It's up to the designers of the survey to control for these things within the survey and/or disclose them. Unfortunately, they haven't.