4 ms·
"The survey sample size was 37 companies from the Reach Capital portfolio. That’s large enough to see patterns"...[citation needed]. n=37 is skimpy.
by ElfinTrousers 4y ago
"The survey sample size was 37 companies from the Reach Capital portfolio. That’s large enough to see patterns"...[citation needed]. n=37 is skimpy.
- blackbear_ 4y agoStatistically speaking, n>=30 is indeed considered a "large" sample, but that is assuming the examples are iid (independent and identically distributed) which is definitely not the case here.
- PeterisP 4y agoIf the effect size is this large, then n=37 is perfectly sufficient to assert that the effect exists; however, the issue is about sample selection, as "n companies from the Reach Capital portfolio" can't be reasonably considered to be a representative sample of all startups, so it can't be assumed that the conclusions can be extrapolated beyond the niche of startups like these which Reach Capital selects to fund, perhaps it depends on location or domain or size or other characteristics which affect what Reach Capital funds.
- invalidname 4y agoMaybe. But there are so many biases that could impact it. Smaller and poorer companies to begin with will have lower revenue and reduce office spend. It's really hard to account for all of these biases.