That’s what I assumed as well, that it was 2% per incident, but I’m having a little trouble parsing those differently:
How is 2% per incident different than 2% per year? I’d interpret both of those statements as ‘on average, given perfect use, a condom will be ineffective at preventing pregnancy in one use out of fifty’.
Well, yes, but what does 2% failure rate per year even mean when it’s presented independent of a number of uses per year? I mean, without knowing what number of average uses were used to calculate “2% failure rate per year”, it seems like somewhat of a misleading statement, as I’m reasonably certain (let’s say at least 90%) that it’s not intended to reflect that condoms become more protective the more chances you have to use them.
I feel like I’m missing something basic here that would let me see why it’s a useful piece of information on its own.
what does 2% failure rate per year even mean when it’s presented independent of a number of uses per year
This is a good observation. You can look up what the average number of uses per year is. If I remember right, I’ve seen some condom efficacy studies include that information.
I feel like I’m missing something basic here that would let me see why it’s a useful piece of information on its own.
You’re not missing anything basic, you’re correctly perceiving ambiguity where ambiguity does exist. Even when information is really important, I’ve found that it’s often been omitted simply because products are marketed to the average person, not to nerdy people like me, and most people don’t want to think as much as I do. For this reason, I’ve found it’s very important to be careful not to assume that the world is doing sensible things or giving me all the information. They’re not just leaving information out, they’re also not being held accountable by a world full of people who think as much as I do. Therefore, they can get away with slapping various nonsense marketing claims and out-of-context data on their boxes without people questioning them.
Thanks for taking a moment to let me know that my comment is appreciated and that this information makes a difference for you. I find that, like Luke says in The Power of Reinforcement, knowing that a behavior of mine has made a difference and is wanted “increases the probability that the behavior will occur again”.
I think LessWrong could really use more positive reinforcement, so I hereby positively reinforce you for showing the humility to positively reinforce.
I agree it’s misleading. That’s why I’d prefer a metric like this, where the precise definitions you use don’t matter if they are the same in both the numerator and the denominator as constant factors would cancel out.
Let’s say, hypothetically speaking, that the average number of uses per year is 100.
A 2% per incident risk will add up to a yearly 50% risk for the average user.*
A 2% per year risk already included 100 uses, so it is still 2% per year.
A 2% per year risk would add up to a 70% chance over the 35 or so years women are fertile and active and a 2% per incident risk would add up to a much, much higher risk, likely resulting in multiple pregnancies.*
* This is only if pure math reflects reality, which it probably doesn’t because there are other factors here like people forgetting important parts of the instructions over time, people getting better at using them over time, or people becoming sloppy about applying them because they’re tired of them or have developed a sense of over-confidence.
No, it requires that the failures of condom usage be independent events from one another. That is to say, that person A using a condom at time B has the same probability of failure as person C using a condom at time D, even if C=A or B=D.
Without knowing more, it is entirely possible that some fraction of men have supersperm which gets through condoms and that they account for all the failures, and those that use condoms but avoid supersperm will never fail. Alternatively it is possible that some group of people keep making the same mistakes and account for all the failures while another group of people use them right and don’t fail.
Math can be used to analyze all of these possibilities.
That’s what I assumed as well, that it was 2% per incident, but I’m having a little trouble parsing those differently:
How is 2% per incident different than 2% per year? I’d interpret both of those statements as ‘on average, given perfect use, a condom will be ineffective at preventing pregnancy in one use out of fifty’.
I guess the average condom user uses more than one per year.
Well, yes, but what does 2% failure rate per year even mean when it’s presented independent of a number of uses per year? I mean, without knowing what number of average uses were used to calculate “2% failure rate per year”, it seems like somewhat of a misleading statement, as I’m reasonably certain (let’s say at least 90%) that it’s not intended to reflect that condoms become more protective the more chances you have to use them.
I feel like I’m missing something basic here that would let me see why it’s a useful piece of information on its own.
This is a good observation. You can look up what the average number of uses per year is. If I remember right, I’ve seen some condom efficacy studies include that information.
You’re not missing anything basic, you’re correctly perceiving ambiguity where ambiguity does exist. Even when information is really important, I’ve found that it’s often been omitted simply because products are marketed to the average person, not to nerdy people like me, and most people don’t want to think as much as I do. For this reason, I’ve found it’s very important to be careful not to assume that the world is doing sensible things or giving me all the information. They’re not just leaving information out, they’re also not being held accountable by a world full of people who think as much as I do. Therefore, they can get away with slapping various nonsense marketing claims and out-of-context data on their boxes without people questioning them.
Yeah, this is good advice in general, and it’s definitely what I was doing wrong this time.
Thanks for taking a moment to let me know that my comment is appreciated and that this information makes a difference for you. I find that, like Luke says in The Power of Reinforcement, knowing that a behavior of mine has made a difference and is wanted “increases the probability that the behavior will occur again”.
I think LessWrong could really use more positive reinforcement, so I hereby positively reinforce you for showing the humility to positively reinforce.
You’re welcome.
I agree it’s misleading. That’s why I’d prefer a metric like this, where the precise definitions you use don’t matter if they are the same in both the numerator and the denominator as constant factors would cancel out.
Here’s how 2% per incident is different:
Let’s say, hypothetically speaking, that the average number of uses per year is 100.
A 2% per incident risk will add up to a yearly 50% risk for the average user.*
A 2% per year risk already included 100 uses, so it is still 2% per year.
A 2% per year risk would add up to a 70% chance over the 35 or so years women are fertile and active and a 2% per incident risk would add up to a much, much higher risk, likely resulting in multiple pregnancies.*
* This is only if pure math reflects reality, which it probably doesn’t because there are other factors here like people forgetting important parts of the instructions over time, people getting better at using them over time, or people becoming sloppy about applying them because they’re tired of them or have developed a sense of over-confidence.
No, it requires that the failures of condom usage be independent events from one another. That is to say, that person A using a condom at time B has the same probability of failure as person C using a condom at time D, even if C=A or B=D.
Without knowing more, it is entirely possible that some fraction of men have supersperm which gets through condoms and that they account for all the failures, and those that use condoms but avoid supersperm will never fail. Alternatively it is possible that some group of people keep making the same mistakes and account for all the failures while another group of people use them right and don’t fail.
Math can be used to analyze all of these possibilities.
I think the figure you’re looking for is 1 - (0.98^100) = 0.87 (assuming no-one gets pregnant twice in the same year).
Well if you have sex once per year, it is not different.
My condolences.