I think those people interpreted the question as “Never left the community” or some such, rather than the site itself.
Also if you add up the LW and LW Meetup numbers they’re almost that, I’d have to play with the data real quick to see how much overlap there is. Eh it’s just a SQL query I’ll do it right now:
sqlite> select count(*) from data where ActiveMemberships_1 = “Yes”;
354
sqlite> select count(*) from data where ActiveMemberships_2 = “Yes”;
216
sqlite> select count(*) from data where ActiveMemberships_1 = “Yes” AND ActiveMemberships_2 = “Yes”;
77
77 / 354 = 0.2175141242937853
77 / 216 = 0.35648148148148145
P(Meetups | LW Site) = 21.75%
P(LW Site | Meetups) = 35.63%
I think that goes a long way towards explaining that particular inconsistency.
However it still leaves a quarter unaccounted for:
I’m trying to reconcile the 353 people on LW with the 668 people who never left. What do you think is up there?
I think those people interpreted the question as “Never left the community” or some such, rather than the site itself.
Also if you add up the LW and LW Meetup numbers they’re almost that, I’d have to play with the data real quick to see how much overlap there is. Eh it’s just a SQL query I’ll do it right now:
sqlite> select count(*) from data where ActiveMemberships_1 = “Yes”;
354
sqlite> select count(*) from data where ActiveMemberships_2 = “Yes”;
216
sqlite> select count(*) from data where ActiveMemberships_1 = “Yes” AND ActiveMemberships_2 = “Yes”;
77
77 / 354 = 0.2175141242937853
77 / 216 = 0.35648148148148145
P(Meetups | LW Site) = 21.75%
P(LW Site | Meetups) = 35.63%
I think that goes a long way towards explaining that particular inconsistency.
However it still leaves a quarter unaccounted for:
(668 - (354 + (216 - (216 * 0.3563)))) / 668 = 0.2619173652694611