(no subject)

[jack]

Ewx's recently made a poll on "number of sexual partners". The graph appeared rather like a poisson distribution[1] but it's hard to actually deduce anything for sure as (a) rounding to the nearest 3 obscures the first part of the hump and (b) most things look vaguely like a poisson anyway.

Statistics is hard. I know a little about the most common distributions, but not really what happens when you start combining them.

The sum of two poisson random variables (even with different means) is another poisson random variable. So I hypothesised that the the number of sexual partners a person has would be a poisson distribution -- if you assume each year they have a poisson-distributed number of partners (eg. mean 1/20 for someone in a stable relationship, 100 for someone very promiscuous), and those means are fixed in advance, it seems the sum to any point in their life will be a poisson.

However, Question 1 is, "What if those means aren't fixed in advance?" Is there a convenient distribution on some appropriate assumptions, or is it just hard?

Question 2 is, supposing each person were a poisson distribution, with means were distributed in a certain way (say, normal distribution), does that give any coherent distribution when you count the number of different people with a number of partners?

Question 3 is, is there any data on this, from Kinsey or anywhere? A cursory search didn't show it. What should ewx's poll show? A hump at 0 and 1? A poisson? Might people's means cluster, if you classed people into different categories, might you see several overlayed poisson curves?

[1] Poisson distribution is what lots of discrete things end up as. If you have a very large binomial distribution, eg. toss 1000 coins with 1/1000 chance of coming up heads, it strarts to look like a poisson. Although what it really measures is when the "time until the next event" is random, ie. exponentially decays.

Comments

[cornute.livejournal.com]

This response is a placeholder for my actual response, which I haven't yet formulated but will as soon as work slows up a bit.

However, I do find this post interesting in a meta sense-- you've just discovered the secret ingredient to a sex post with no responses. Maths!

[d37373.livejournal.com]

Poisson is good for events occurring at irregular intervals (say, finding somebody you'd want to hook up with), but the distribution is going to be skewed by the length of time during which a person is not looking, e.g. monogamous and in a relationship.

Also, polygamy will be overrepresented, because of the n²relationships in group of n people.

You're obviously going to have somewhat different numbers for different ages, because the count can only increase, not decrease. Whether this affects the shape of the distribution or not is another question, and one which I'm sure statistics can answer apart from any situational information.

[sunflowerinrain.livejournal.com]

Statistics depends on having a sufficient quantity of really good data, but also on analysing them with a huge amount of background knowledge. See tig_b for details.

Background knowledge to this particular problem would include that people go through different phases wrt number_of_sexual_partners. For example, a particular life-event might trigger either celibacy or promiscuity. Many people are promiscuous for a short period in their life: possibly before settling into a stable partnership, but it can also happen that someone who is celibate before marriage and then divorces goes seriously partying and suddenly their total rises from 1 to lots.

[simont]

"number of sexual partners"
[...]
rounding to the nearest 3 obscures the first part of the hump

Was that deliberate? If not, can I be the first to post a puerile snigger? *snigger*

[cartesiandaemon.livejournal.com]

you've just discovered the secret ingredient to a sex post with no responses. Maths!

Although you seem to underestimate my friends-list, there aren't *lots* of responses, but statistics definitely seems as provocative as sex :)

[cartesiandaemon.livejournal.com]

ROFL. No, entirely accidently, thanks for spotting it :)

[cartesiandaemon.livejournal.com]

I was going to ask "if you were a poisson distribution, what would the mean of your sexual partners be" and get feedback but that's not really something you can ask, you have to ask in aggregate...

[megamole.livejournal.com]

I wish you hadn't used "hump" in that context.

[megamole.livejournal.com]

Darn, Simon got there first.

[cartesiandaemon.livejournal.com]

So to speak :)