2

Why don't epidemiologists ask people when they had sex? Why do they estimate time of conception from the birth date? This seems to introduce unnecessary error. Wouldn't it be better to record the time you had sex in some sort of database (assuming it was the time it resulted in pregnancy)?

Note that fertilization of the egg by sperm can happen without sex.

Fomite
  • 23,134
guest2323
  • 21
  • 1
    Doesn't everybody maintain such a database on their PC? :-) – whuber Mar 05 '13 at 03:25
  • @whuber: I'm not following you. Meaning that non-statisticians would most probably not be meticulous enough to record when they had sex whereas statisticians would be? – guest2323 Mar 05 '13 at 03:27
  • 4
    I seriously doubt anybody, except a few pathological souls, would be so meticulous, statisticians least of all (because they know better). Think about it: if you need to be recording these times and dates, then you're likely having sex a lot more than once a month (it happens to a few lucky couples :-), in which case the time of conception is still indeterminate, so what's the use in having the record? Statisticians would also be aware that gestation is variable, so that a full term might be 39 weeks for some women, 41 for others. In light of that a precise record is superfluous. – whuber Mar 05 '13 at 03:33
  • It appears you have two accounts. If you merge them (at http://stats.stackexchange.com/help/user-merge) you will be able freely to comment in this thread and to edit your question if you wish. – whuber Mar 05 '13 at 04:48
  • I think your "assuming it was the time it resulted in pregnancy" is a big assumption that in most cases we'd have no way at all to test. – curious_cat Mar 05 '13 at 05:26

2 Answers2

3

A number of things to consider:

  • Sometimes we do. There are, on occasion, studies that ask you to keep what is, for lack of a better description, a diary of sexual activity.
  • We don't just back calculate from birth date. There are a number of methods of estimating date of conception - the one I see far more frequently than your suggested method is using the last known menstrual period of the mother. But there are extremely precise methods we can use - one study had women submit a staggering number of urine samples to detect pregnancy hormones, for example, which give extremely precise estimates.
  • Your suggested approach is neither free of error, nor necessarily better. There are two reasons for this - first, while sex is a necessary component of conception, it isn't sufficient. "Had sex" doesn't indicate conception occurred. If its a once in a blue moon event, then sure it might work, but humans have sex for many reasons besides reproduction. Which means:

A. The timing of your estimate is still imprecise, and will probably be based informally on birth date. "What was I doing 9 months ago...ah, after that party at the Hendersons..."
B. If a couple is frequently having sex, you're essentially guessing from a sample.
C. Many subjects won't recall, and there may be serious differences in the recall rates for people based on important pieces of information (the non-random error another poster mentions). For example, couples just having sex might not remember with any precision, yet couples trying to have a baby might be tracking these things quite carefully. Random error, like the kind introduced by back calculating from a birthdate is irksome. Non-random error is bad.

Fomite
  • 23,134
2

There is a wealth of gestational age and birth weight population data and the prediction is less error-prone that one may expect. Just like growth reference chart for children, there are growth reference chart for fetus as well.

In addition, when it comes to interview, it can be burdensome for respondents to disclose when they have sex, mostly due to memory fading or the pregnancy being socially marginalized (especially in situation like teenage or underage moms, out-of-wedlock conception, etc.) Some people have sex once a while, some have sex multiple times a day, each of them would be subjected to different memory biases. And especially the condition you said:

assuming it was the time it resulted in pregnancy,

for those who have a lot of frequent sex, how can you even tell? It's not like the woman's nose will blink like Rudolph's if the sperm made it to her egg.

Keeping the estimation of conception date in-house can skip many awkward moments, and perhaps will result in a higher response rate and a much trust-worthy assessment.

Penguin_Knight
  • 11,988
  • @PenguinKnight: I guess it would be easier if you have sex once and then pregnancy results. If you had sex a lot, it would be difficult. But for people who have functional sex, I think it would be easier to approximate the time of conception (especially if you don't need to know the exact time of conception). For example, having sex once on Tuesday which results in pregnancy means that the conception date was on Tuesday at some time. We don't need to know the exact time to the millisecond. By functional sex I mean that the people have sex for the purpose of having a baby (and maybe pleasure is –  Mar 05 '13 at 03:48
  • 1
    Correct, but once again you allude to the fact that the estimate is only good for some people, but not the others. We epidemiologists love to keep the error non-systematic (aka random). If we estimate the conception date with one reference data, error can be assumed to be equal in all sexual lifestyles; if we asked them for the date of having sex, different groups will provide answers with various accuracies, and that can potentially induce bias, as the relationship is now confounded by the frequency of sex, which can either be observed or worse, unobserved (aka we never asked the mother.) – Penguin_Knight Mar 05 '13 at 03:55