3 ms·
Not that I am a mathematician, but I am pretty sure that you can't actually calculate how many 40+ people there are just from these values since you can't assum
by danieltillett 11y ago
Not that I am a mathematician, but I am pretty sure that you can't actually calculate how many 40+ people there are just from these values since you can't assume the distribution is normal.
- nl 11y agoWell, it's clearly not normal because it bound at the low end. But assuming a normal distribution shape for the high side is probably a reasonable first guess. Put it like this: I'd say that the average being 29 means that around 80% of founders funded by YC are below 40.
- danieltillett 11y agoAssuming a normal distribution is always a reasonable first step, but when dealing with small numbers in your distribution you have to be really careful as you can go massively wrong with this assumption.
- eterm 11y agoThat's true, but you can ignore distributions, and assert that they must accept more people under 29 than over 40.
- danieltillett 11y agoTrue, but this doesn't help you too much in calculating how many are over 40.
- Xophmeister 11y agoThe central limit theorem might have something to say about that...
- dagw 11y agoI doubt they have funded a Sufficiently Large Number of people.
- danieltillett 11y agoExactly.
- nl 11y agoIt's well over 1600 people funded. I'd find it surprising if we didn't see a skewed normal distribution (skewed because of the lower bound).
- Florin_Andrei 11y agoYeah, but I'm not sure about the "independent random variables" part. Specifically, the "independent" qualifier. This is not a bunch of uncorrelated uranium decay random number generators.
- dagw 11y agoWell we can set hard upper and lower bounds on the ages of people they fund (let's call them 18 and 80). And if we further assume that 18-22 is more common than 65-80, that gives us fair bit of information to start building a likely distribution.
- danieltillett 11y agoYes we can build a model, but given we don't know how many people have been funded we can't really be very precise.
- nl 11y agoIt's the distribution that is interesting. (I'm actually pretty sure there is enough information about on how many companies YC is funded and the average team size to have a decent go at estimating the absolute numbers too, but that is less interesting for the purposes of this discussion) Edit: (since you seem interested in the numbers) based on [1] and [2] we can guess lower bounds of (a) 550 companies funded, (b) 2 founders per company and taking my previous "(c)80% less than 40 estimate" (based on a normal distribution) that gives around 220 older than 40. Given than we don't know (a),(b) or the standard deviation needed to accurately estimate (c) it is actually more useful to talk about the estimate (c) than to increase the inaccuracy by using the numbers (a) and (b). Edit 2: [3] gives us more accurate numbers. 842 companies funded, max age of founder 66 (I guessed a max of 70). Using that to adjust the standard deviation, and sticking with the (b)=2 estimate above we get around 294 founders 40 or older. Call it 300 in round numbers. To verify this, someone could run the numbers to check if this gives the average age (30.27) along with the median of 29 (although guessing the shape of the lower-side curve is probably challenging). [1] http://yclist.com/ http://yclist.com/ [2] http://www.quora.com/What-are-the-average-demographics-of-those-accepted-into-the-Y-Combinator-program http://www.quora.com/What-are-the-average-demographics-of-th... [3] http://blog.ycombinator.com/yc-stats-winter-2015 http://blog.ycombinator.com/yc-stats-winter-2015
- danieltillett 11y agoGiven that the actual value is known we don't really need to calculate anything, just get those that know to tell us. I suspect that the distribution is approximately normal and that your calculations are close to the true value. From my perspective the really interesting thing to know is how close the true numbers are to expected. This will tell us something useful about the selection process.