5 ms·

It’s a weakly supported clickbait article. A may survey of 1000 people reported that 39% would consider quitting if their work didn’t provide flexibility. I’m

by wepple 5y ago

It’s a weakly supported clickbait article. A may survey of 1000 people reported that 39% would consider quitting if their work didn’t provide flexibility. I’m no statistician but the red flags are n=1000 and “would consider”. This is a hugely hit button issue right now, great for clicks.

elzbardico 5y ago
Right now for software engineers, quitting is not that complicated. Easy to find another offer. Heck, recruiters can't leave us alone a single day.
aninteger 5y ago
Clickbait yes, but it's more than that. It's a momentum/filler piece. NPR's Marketplace ran a short segment on this last week (although I can't find the link), now Bloomberg, coming up MSNBC and CNN(?).
wgerard 5y ago
> red flags are n=1000 It's been awhile since I've taken stats, but people semi-frequently overestimate the necessary sample size needed for even a 99% confidence level. Depends on the standard deviation for sure, but some quick napkin math suggests that a sample size of ~700 people is more than enough to draw inferences about the entire western hemisphere at a 99% confidence level.
dredmorbius 5y ago
For a single question survey, 1,000 is a more than sufficiently large survey. Standards are typically for 300+ given random sampling. Reasons for enlarging samples have less to do with accuracy than with resolution. If you plan on looking at specific cross-tabulations or sub-populations, you need a sufficient sample size (typically n>30 for large-sample statistics) to draw inferences. Error decreases with the square root of sample size. To halve error you must square the sample. A 1,000-element sample has half the error of a 31-element sample. To halve error again would require a 1,000,000 element sample. "Would consider" is a fair catch, though here the question is how that compares against previous measures / trend.

by wepple 5y ago

elzbardico 5y ago

Right now for software engineers, quitting is not that complicated. Easy to find another offer. Heck, recruiters can't leave us alone a single day.

aninteger 5y ago

Clickbait yes, but it's more than that. It's a momentum/filler piece. NPR's Marketplace ran a short segment on this last week (although I can't find the link), now Bloomberg, coming up MSNBC and CNN(?).

wgerard 5y ago

> red flags are n=1000 It's been awhile since I've taken stats, but people semi-frequently overestimate the necessary sample size needed for even a 99% confidence level. Depends on the standard deviation for sure, but some quick napkin math suggests that a sample size of ~700 people is more than enough to draw inferences about the entire western hemisphere at a 99% confidence level.

dredmorbius 5y ago

For a single question survey, 1,000 is a more than sufficiently large survey. Standards are typically for 300+ given random sampling. Reasons for enlarging samples have less to do with accuracy than with resolution. If you plan on looking at specific cross-tabulations or sub-populations, you need a sufficient sample size (typically n>30 for large-sample statistics) to draw inferences. Error decreases with the square root of sample size. To halve error you must square the sample. A 1,000-element sample has half the error of a 31-element sample. To halve error again would require a 1,000,000 element sample. "Would consider" is a fair catch, though here the question is how that compares against previous measures / trend.