What are the odds?
March 23, 2011 12:16 PM Subscribe
What are the odds, really, of having identical triplets?
Okay, I'm not really asking for the odds, but I do want to be able to dispute a statistic that is often quoted in the media and by other people.
It is often said that the odds of having identical triplets are 1:200 million. Which to me sounds ridiculous, but I'm not sure if my reasoning is sound.
We have identical triplets, born in December of 2007. I know of several other sets, also born in 2007. At least 3 (one set in 4/2007, one in 8/2007, and another set in 12/2007.) These are only ones I've met online. I'm assuming there are more out there.
According to the CDC, there were 4,317,119 births in 2007. (I don't know if they count a multiples birth as one, or if it's one per kid, but I don't think that really matters for the purposes of this question.) Four of those, at least, were ID triplets. That's more like 1:1,000,000, and those are only the sets I know. So it's likely to be even less than that. Right?
I'm just annoyed by reading that 1:200 million statistic, and would like to be able to refute it intelligently. (I'm just posting information from 2007 since that's what I have, but I assume the birth rates are the same every year.)
As you can tell from a previous question, I don't really understand probability and statistics all that well. So use layman's terms, if you don't mind.
Okay, I'm not really asking for the odds, but I do want to be able to dispute a statistic that is often quoted in the media and by other people.
It is often said that the odds of having identical triplets are 1:200 million. Which to me sounds ridiculous, but I'm not sure if my reasoning is sound.
We have identical triplets, born in December of 2007. I know of several other sets, also born in 2007. At least 3 (one set in 4/2007, one in 8/2007, and another set in 12/2007.) These are only ones I've met online. I'm assuming there are more out there.
According to the CDC, there were 4,317,119 births in 2007. (I don't know if they count a multiples birth as one, or if it's one per kid, but I don't think that really matters for the purposes of this question.) Four of those, at least, were ID triplets. That's more like 1:1,000,000, and those are only the sets I know. So it's likely to be even less than that. Right?
I'm just annoyed by reading that 1:200 million statistic, and would like to be able to refute it intelligently. (I'm just posting information from 2007 since that's what I have, but I assume the birth rates are the same every year.)
As you can tell from a previous question, I don't really understand probability and statistics all that well. So use layman's terms, if you don't mind.
Here's two data points:
The National Center for Health Statistics says there were 294 sets of triplets born in the United States in 1944 out of a total of 2,794,800 births.
There is no record of how many of the 294 were identical triples. But Dr. Louis G. Keith of the Northwestern University Medical School, who has written extensively on this subject, believes most of those 294 were probably identical.
With fertility treatments affecting birth rates now, in 2002 there were 6,898 sets of triplets out of 4,021,726 births in the U.S. and again, there is no record of how many of these were identical. Fertility treatments do not cause identical triplets to be born as they utilize individual eggs for their zygotes. Identical triplets are mono-zygotic (a single zygote which splits into three). So with more triplets being born and the rate of identical triplets staying the same, identical triplets are more rare now than they were back in 1944.
The odds of conceiving "spontaneous" triplets is about 1 in 8,100. In the early '80s, triplets occurred 1 in every 6,400. Today, odds are that 1 in every 1,300 are triplets. Over the last decade, increases in the number of triplet births averaged 11 percent a year.
The 1 in 200 million quote seems completely fabricated, if "most" triplets born pre-fertility-drugs were identical. However, keep in mind triplets born in 2007 are still fairly young; they might not look as identical as they get older. We'll need DNA tests to be sure.
posted by gerryblog at 12:28 PM on March 23, 2011 [3 favorites]
The National Center for Health Statistics says there were 294 sets of triplets born in the United States in 1944 out of a total of 2,794,800 births.
There is no record of how many of the 294 were identical triples. But Dr. Louis G. Keith of the Northwestern University Medical School, who has written extensively on this subject, believes most of those 294 were probably identical.
With fertility treatments affecting birth rates now, in 2002 there were 6,898 sets of triplets out of 4,021,726 births in the U.S. and again, there is no record of how many of these were identical. Fertility treatments do not cause identical triplets to be born as they utilize individual eggs for their zygotes. Identical triplets are mono-zygotic (a single zygote which splits into three). So with more triplets being born and the rate of identical triplets staying the same, identical triplets are more rare now than they were back in 1944.
The odds of conceiving "spontaneous" triplets is about 1 in 8,100. In the early '80s, triplets occurred 1 in every 6,400. Today, odds are that 1 in every 1,300 are triplets. Over the last decade, increases in the number of triplet births averaged 11 percent a year.
The 1 in 200 million quote seems completely fabricated, if "most" triplets born pre-fertility-drugs were identical. However, keep in mind triplets born in 2007 are still fairly young; they might not look as identical as they get older. We'll need DNA tests to be sure.
posted by gerryblog at 12:28 PM on March 23, 2011 [3 favorites]
Are they genetically identical or just looks REALLY REALLY similar? Mary-Kate and Ashley Olsen, for example, are not genetically identical (I read somewhere in the last 2 decades.)
posted by k8t at 12:29 PM on March 23, 2011
posted by k8t at 12:29 PM on March 23, 2011
Best answer: According to this page identical triplets occur in 1 in 62500 pregnancies. It looks truthy and facty to me, though who knows.
posted by PercussivePaul at 12:35 PM on March 23, 2011
posted by PercussivePaul at 12:35 PM on March 23, 2011
Best answer: Based on UK statistics (2009, Office of National Statistics, General Registry Office Scotland and GRO Northern Ireland), the rate of any twin is 16 in 1000 births, and the rate of identical twins is 1/3 that, or figure 5 in 1000 births.
I think the formation of an identical triplet would require first the formation of an identical twin, and then one of those twins to form another identical twin. So my estimate would be 5 in one million.
posted by babbageboole at 12:37 PM on March 23, 2011
I think the formation of an identical triplet would require first the formation of an identical twin, and then one of those twins to form another identical twin. So my estimate would be 5 in one million.
posted by babbageboole at 12:37 PM on March 23, 2011
And, of course, I can't multiply. I meant 25 in one million, or 1 in 40,000. That seems a bit high, though.
posted by babbageboole at 12:38 PM on March 23, 2011
posted by babbageboole at 12:38 PM on March 23, 2011
Response by poster: gerryblog - that's a great article!
k8 - send me your address and I'll send you one.
PercussivePaul, I've always liked that statistic. It seems more realistic to me. That's the one I generally tell people when they ask how rare it is. I think I must have gotten it from that same page.
Babbageboole, I like the way you came up with that. That makes sense to me.
posted by pyjammy at 12:47 PM on March 23, 2011 [2 favorites]
k8 - send me your address and I'll send you one.
PercussivePaul, I've always liked that statistic. It seems more realistic to me. That's the one I generally tell people when they ask how rare it is. I think I must have gotten it from that same page.
Babbageboole, I like the way you came up with that. That makes sense to me.
posted by pyjammy at 12:47 PM on March 23, 2011 [2 favorites]
If the odds of having identical triplets were one in 200 million, there would be about thirty-five sets of them in the world -- seven billion divided by 200 million. This seems unreasonably low.
babbageboole: there's something called "Hellin's Law" which is a rule of thumb for the incidence of all multiple births, not just identical ones; it states that the frequency of twins is one in 89, of triplets is one in 892, and so on. This agrees with your intuition but I don't know enough about the mechanism to say whether it would apply specifically to identical multiple births.
posted by madcaptenor at 12:49 PM on March 23, 2011
babbageboole: there's something called "Hellin's Law" which is a rule of thumb for the incidence of all multiple births, not just identical ones; it states that the frequency of twins is one in 89, of triplets is one in 892, and so on. This agrees with your intuition but I don't know enough about the mechanism to say whether it would apply specifically to identical multiple births.
posted by madcaptenor at 12:49 PM on March 23, 2011
Response by poster: 23skidoo - really? What about incorrect statistics, like the one my question is about?
posted by pyjammy at 12:54 PM on March 23, 2011
posted by pyjammy at 12:54 PM on March 23, 2011
Best answer: I'm seeing statistic similar to babbage's: worldwide, the rate for monozygotic twins is 4 in 1000, which (in theory) calculates out to 16:1000000 for triplets, or 1:62500. As PercussivePaul points out, this is a number that a lot of places cite.
posted by specialagentwebb at 1:17 PM on March 23, 2011
posted by specialagentwebb at 1:17 PM on March 23, 2011
Response by poster: Ah well, I did say I'm not great at understanding statistics.
posted by pyjammy at 1:24 PM on March 23, 2011
posted by pyjammy at 1:24 PM on March 23, 2011
Best answer: Then, you need to look at data over a span of several decades, not just one year.
Then, you can calculate the correct odds.
Not when the alleged probability is that far off from the observed figure. If the alleged probability were, say, 1 in 2 million, while the observed rate in 2007 was about 1 in 1 million, you're right that you'd need to collect more data.
However, if the actual probability of a birth being identical triplets is 1 in 200,000,000, then the probability of 4 or more births out of 4317119 being identical triplets is on the order of 10-8.
(Calculated as 1 minus the probability of having 0-3 identical triplets out of those 4317119, which in turn is sum{i=0 to 3} pi(1-p)N-1C[N,i], where p=1/200000000, N=4317119, and C[x,y] is the binomial coefficent.)
You might object that that would be the probability in a randomly chosen year, and there may be some selection bias—no one questions the lack of identical triplets in years when that never happens. Which would be a valid objection. So the probability of 4 identical triplets being born in the same year any time in US history would be less than 2x10-6 (probably much less, given that most of those years would have had far fewer than 4317119 births), if the cited probability were correct.
We can quite confidently reject the hypothesis that the probability of identical triplets is 1 in 200 million per birth in the United States without the need to collect any further data.
What you're suggesting is like saying "I flipped a coin 10 times, and did not get 5 heads and 5 tails, therefore the 50/50 odds of coinflips is wrong."
No, what pyjammy is suggesting is like saying "I flipped a coin 32 times and it came up heads 31 times and tails once, therefore this coin does not have a 0.5 probability of coming up heads on a single flip."
posted by DevilsAdvocate at 1:34 PM on March 23, 2011 [3 favorites]
Then, you can calculate the correct odds.
Not when the alleged probability is that far off from the observed figure. If the alleged probability were, say, 1 in 2 million, while the observed rate in 2007 was about 1 in 1 million, you're right that you'd need to collect more data.
However, if the actual probability of a birth being identical triplets is 1 in 200,000,000, then the probability of 4 or more births out of 4317119 being identical triplets is on the order of 10-8.
(Calculated as 1 minus the probability of having 0-3 identical triplets out of those 4317119, which in turn is sum{i=0 to 3} pi(1-p)N-1C[N,i], where p=1/200000000, N=4317119, and C[x,y] is the binomial coefficent.)
You might object that that would be the probability in a randomly chosen year, and there may be some selection bias—no one questions the lack of identical triplets in years when that never happens. Which would be a valid objection. So the probability of 4 identical triplets being born in the same year any time in US history would be less than 2x10-6 (probably much less, given that most of those years would have had far fewer than 4317119 births), if the cited probability were correct.
We can quite confidently reject the hypothesis that the probability of identical triplets is 1 in 200 million per birth in the United States without the need to collect any further data.
What you're suggesting is like saying "I flipped a coin 10 times, and did not get 5 heads and 5 tails, therefore the 50/50 odds of coinflips is wrong."
No, what pyjammy is suggesting is like saying "I flipped a coin 32 times and it came up heads 31 times and tails once, therefore this coin does not have a 0.5 probability of coming up heads on a single flip."
posted by DevilsAdvocate at 1:34 PM on March 23, 2011 [3 favorites]
Well, just remember that there are lies, damn lies, and statistics. Also, 23skidoo... statistics isn't about truth, it's about estimation. If you want to estimate something random, you have to measure it, and then assume a particular random process, and even then statistics will tell you how likely it is that your estimate matches your assumption.
"Makes sense" and "seems realistic" are just intuitive methods of determining the likelihood of your estimate matching an underlying process, just as statistical hypothesis testing is a mathematical method.
posted by babbageboole at 1:36 PM on March 23, 2011
"Makes sense" and "seems realistic" are just intuitive methods of determining the likelihood of your estimate matching an underlying process, just as statistical hypothesis testing is a mathematical method.
posted by babbageboole at 1:36 PM on March 23, 2011
Does this count? I don't think so.
I know one set of triplets: the two girls are identical and there's also a boy (not identical). Is it worth mentioning they were born on February 29?
Other than the leap year business, is this more common?
posted by crankyrogalsky at 2:03 PM on March 23, 2011
I know one set of triplets: the two girls are identical and there's also a boy (not identical). Is it worth mentioning they were born on February 29?
Other than the leap year business, is this more common?
posted by crankyrogalsky at 2:03 PM on March 23, 2011
that's still not a reason to promote "making sense" and "seeming realistic" as good things to do when evaluating statistics, because people suck at estimating
What? Really? If we do a calculation and come up with a figure that the average person owns 63 cars, we shouldn't ask ourselves if that makes sense as a means to evaluate whether or not the calculation was done correctly? How do you propose we detect when statistical things were figured incorrectly? The statistic of 1 in 200 million births being triplets defies expectation because (it seems) most people know a set of triplets, but we know most people don't know 200 million people. It absolutely warrants further investigation on that basis.
posted by 0xFCAF at 2:20 PM on March 23, 2011
What? Really? If we do a calculation and come up with a figure that the average person owns 63 cars, we shouldn't ask ourselves if that makes sense as a means to evaluate whether or not the calculation was done correctly? How do you propose we detect when statistical things were figured incorrectly? The statistic of 1 in 200 million births being triplets defies expectation because (it seems) most people know a set of triplets, but we know most people don't know 200 million people. It absolutely warrants further investigation on that basis.
posted by 0xFCAF at 2:20 PM on March 23, 2011
Are they genetically identical or just looks REALLY REALLY similar? Mary-Kate and Ashley Olsen, for example, are not genetically identical (I read somewhere in the last 2 decades.)
The New York Times mentioned it in their story about the Olsen twins in this article. Their younger sister also looks very similar to them.
posted by anniecat at 2:49 PM on March 23, 2011 [1 favorite]
The New York Times mentioned it in their story about the Olsen twins in this article. Their younger sister also looks very similar to them.
posted by anniecat at 2:49 PM on March 23, 2011 [1 favorite]
Response by poster: crankyrogalsky - that is not all that uncommon, actually. "Pair and a spare" is sometimes what it's called. The birthday, however, is certainly unusual! I mean, not statistically, because I'm only guessing.
posted by pyjammy at 3:18 PM on March 23, 2011
posted by pyjammy at 3:18 PM on March 23, 2011
I would guess that most "identical" twins and triplets are not genetically identical. Twins run in my family quite a bit and I don't think any of them are genetic twins (based on them saying/knowing definitively they're not) although most of them do look very alike. My mom and her twin sister look identical to non family but they're not.
posted by fshgrl at 3:38 PM on March 23, 2011
posted by fshgrl at 3:38 PM on March 23, 2011
FWIW, unless any of your acquaintances live in Portland, I know of YET ANOTHER set of identical triplets born in 2007.
posted by KathrynT at 10:10 PM on March 23, 2011
posted by KathrynT at 10:10 PM on March 23, 2011
Response by poster: fshgrl - only fraternal twins "run" in families, identical twins/triplets are flukes. I am only counting people who knew for sure that their triplets are identical. I had mine DNA tested, but when you have 20 ultrasounds in your pregnancy, it's easy to know that they are indeed monochorionic (i.e., identical.)
kathryn - interesting! I don't think I know that set. (The ones I know all have blogs or are online in some way. If they do, you can message me.)
posted by pyjammy at 4:16 AM on March 24, 2011
kathryn - interesting! I don't think I know that set. (The ones I know all have blogs or are online in some way. If they do, you can message me.)
posted by pyjammy at 4:16 AM on March 24, 2011
This thread is closed to new comments.
Then, you need to look at data over a span of several decades, not just one year.
Then, you can calculate the correct odds.
What you're suggesting is like saying "I flipped a coin 10 times, and did not get 5 heads and 5 tails, therefore the 50/50 odds of coinflips is wrong."
posted by k5.user at 12:24 PM on March 23, 2011