HIV statistics and do people actually understand them.

[Ts RedVeX]

flighty2, I don't think there is any point in bringing theory of conditional probability into this, as whoever came up with the 0.27% obviously had no idea of it. Look at these values:


Receptive anal sex amongst gay men, partner unknown status: gu=0.27% (1:370)
Receptive anal sex amongst gay men, partner HIV positive g=0.82% (1:123)


Perhaps you have not noticed it, but g=3gu (more less) That means the probability of encountering an HIV-positive homosexual man (let's assume that is what they mean by "gay") is 1/3 (every third gay man has HIV). I remember that i took that 4% of the world population is homosexual. If half of the population are HIV positive males out of whom every third has HIV, then that leaves us with 50666666 HIV-positive homosexual men. Considering that the other stats from https://www.hiv.gov/hiv-basics/overv...bal-statistics (only 36mln people around the world know they are HIV-positive and in fact maybe there are 40% more), the 56,5 mln HIV- positive peeps is already a bit weird as it is the 36M+40% already. That means there are no women with HIV, or heterosexual men with HIV at all. Don't you think that is "a bit iffy" - to say the least? This is the reason why I think this whole table is some sort of anti-homosexual propaganda.

[trish]

Let P(C|R^G) denote one’s conditional probability of contracting HIV from a single anal receptive encounter with a gay partner. This then is what you’ve been calling gu. So

P(C|R^G) = 0.0027

Let P(C|R^G^H) denote one’s conditional probability of contracting HIV from a single anal receptive encounter with a gay partner with aids. This is your g. So

P(C|R^G^H) = 0.0082.

The point of using this standard notation from probability theory is that now we can readily use the definition of conditional probability in each of these cases, namely

P(C^R^G)/P(R^G) = 0.0027; and

P(C^R^G^H)/P(R^G^H) = 0.0082.

RedVex notes that the ratio is approximately 3; i.e.

[P(C^R^G^H)/P(R^G^H)]/[P(C^R^G)/P(R^G)] = 3.

However, this does NOT reduce to

P(H^G)/P(G) = 1/3;

i.e. the chart does NOT imply one gay person out of every three has the HIV virus. The bullshit artist here is not HIV.Gov.

[Ts RedVeX]

Frankly, I am not sure what you are writing about Trish. What are the R, G, H, C events?

[KelliBlueEyes]

Isn't a risk matrix constructed of not only the probability of a bad outcome, but also the severity of the outcome should a bad outcome be experienced?

I'd say in the ongoing example the worst possible outcome is cotracting hiv. Even with the mitigating factors such as recent medical developments it's still prudent to limit one's risk. Unless one is ok with the worst possible outcome.

[trish]

Frankly, I am not sure what you are writing about Trish. What are the R, G, H, C events?

Let P(C|R^G) denote one’s conditional probability of Contracting HIV from a single anal Receptive encounter with a Gay partner. This then is what you’ve been calling gu. So

P(C|R^G) = 0.0027

Let P(C|R^G^H) denote one’s conditional probability of contracting HIV from a single anal receptive encounter with a gay partner carrying the HIV virus. This is your g. So

P(C|R^G^H) = 0.0082.

Sorry, I hoped it was clear from the context. Above I highlighted the letters denoting the corresponding events.

[Ts RedVeX]

Since we are only considering cases of receptive anal sex with gay men, why do you even have R and G events in your equations? Those two are certain events.

[trish]

We are considering cases where one has anal receptive sex (that's R) with a partner who is gay (that's G). We're also considering cases where the partner is a carrier of the HIV versus (that's H) and cases where you (the generic you) contract the virus from the encounter (that's C). You pretend to have deduced that P(H|G) = 1/3 from the listed values of P(C|R^G) and P(C|R^G^H). However the value of P(H|G) cannot be determined simply by taking the ratio of the latter two quantities.

[filghy2]

flighty2, I don't think there is any point in bringing theory of conditional probability into this, as whoever came up with the 0.27% obviously had no idea of it. Look at these values:


Receptive anal sex amongst gay men, partner unknown status: gu=0.27% (1:370)
Receptive anal sex amongst gay men, partner HIV positive g=0.82% (1:123)


Perhaps you have not noticed it, but g=3gu (more less) That means the probability of encountering an HIV-positive homosexual man (let's assume that is what they mean by "gay") is 1/3 (every third gay man has HIV). I remember that i took that 4% of the world population is homosexual. If half of the population are HIV positive males out of whom every third has HIV, then that leaves us with 50666666 HIV-positive homosexual men. Considering that the other stats from https://www.hiv.gov/hiv-basics/overv...bal-statistics (only 36mln people around the world know they are HIV-positive and in fact maybe there are 40% more), the 56,5 mln HIV- positive peeps is already a bit weird as it is the 36M+40% already. That means there are no women with HIV, or heterosexual men with HIV at all. Don't you think that is "a bit iffy" - to say the least? This is the reason why I think this whole table is some sort of anti-homosexual propaganda.

The point you seem to be determined to miss is that you cannot have unprotected anal sex with a random client. You can only do so with those who are willing to have unprotected anal sex with a stranger (ie a risk-taker). That is my point about conditional probability. Your calculation assumes that the 0.27% probability is conditional only on the person being gay: in fact, it is conditional on the person being gay and willing to have unprotected sex with strangers.

[Ts RedVeX]

I am only analysing the data given in the tables I gave links to at the beginning of this thread. You can condition your probability on whether or not you take a massive dump or just a small one in the morning, if you know how that affects the probability of contracting HIV. Both values I used refer to unprotected receptive anal sex with a gay man, and that implicates we are both taking the risk. The only difference is that in one case he has HIV for sure, in the other - it is not certain that he has HIV. You are making a mountain out of a molehill. If only you had done that while your global warming research, then maybe we'd not have to worry about our CO2 emissions because it would turn out there are also other important causes for climate change...

If you want to bring some more data in, e.g. on how farting increases the probability, then go ahead. There is not much left to discuss here, really. People generally do not understand stats, they do not trust them, do not know how to read them and what they actually mean, in my opinion.

[Wendy Summers]

If you want to bring some more data in, e.g. on how farting increases the probability, then go ahead. There is not much left to discuss here, really. People generally do not understand stats, they do not trust them, do not know how to read them and what they actually mean, in my opinion.

While I generally agree with the idea that most folks don't get statistics, I don't think that's what's happening from most people in this thread.

From an abstract standpoint, your math may be a reasonable back of the envelope calculation of risk across the entire population. I don't think the push back you are getting here is driven by lack of understanding of your calculation, but instead driven by the debate of whether your calculation is actually useful. As others have noted, your inputs are too simplified for this to have a practical purpose. That's kind of the key here. While math oriented types love digging into numbers like this, the rest of the world evaluates results with the simple question: Does this help me make a better decision?

If your calculation doesn't help people make a better decision, they will reject your model. Many of the suggestions for missing inputs in this thread are aimed at making it a useful tool.