This virology [blog] discusses estimates of the number of mammalian viruses and the costs of 'discovering' 85% of them.

My question is whether this is not a forlorn hope. The ".632 rule" in statistics says roughly that as we approach n random samples from a large population of size n we will see only about 62.3 per cent of the population. After a point we would begin to see the same viruses again and again. This argument is probably stronger with respect to marine viruses, as they may be even more numerous and sampling at depths could be difficult.

I wonder if someone with experience sampling small organisms from large populations or familiarity with the literature has an idea about the plausibility of seeing 85% of a large population, such as mammalian viruses, by sampling?

If the samples are really random (in some sense) and the population is large, a few computer simulations reveal the power of this rule of thumb.

Edit in response to comment/question:

Like any such estimate, the blog's estimate of 3.6 million is a guess. I am not reproducing the calculation here because it's just speculation.

Having said that, if 3.6 million were correct, we would have to draw about 6.8 million samples to find 85% of the existing viruses, 8.2 million to find 90%, and so on. As the blog notes, the PCR approach detects viruses similar to those we know, so the sampling is worse than random. If we look at the difficulty of finding/using a method that has a reasonable chance of capturing any of the extant viruses, the number 3.6 million looks very big.

  • $\begingroup$ How did they know/estimate the total no. of viruses ? $\endgroup$
    – biogirl
    Jun 8, 2014 at 5:07
  • $\begingroup$ Is there an article you could refer me to for the .623 rule? I was very interested and would like to know more and also could you elaborate how you made the calculations in your edit. Thanks $\endgroup$ Jun 8, 2014 at 13:47
  • 1
    $\begingroup$ Excellent page, I found the R code on the page very helpful. Many thanks $\endgroup$ Jun 8, 2014 at 14:32
  • 1
    $\begingroup$ Regarding the question, I guess it is possible to reach the %85 based on the formula however I think there are two major problems that could affect the percentage. First most of the viruses collected for analysis is probably not collected/sampled at random as (unfortunately) samples are collected when they are of interest e.g disease causing viruses etc. Second is that there are perhaps many places, which have not yet been sampled due to inaccessibility etc. So in short I think it is possible but there are barriers but these are just my thoughts and I could be totally wrong! $\endgroup$ Jun 10, 2014 at 16:35
  • 1
    $\begingroup$ @Bez: Yes I think this is all true. And maybe the biggest barrier is the numbers. How to do 7 million random samples. I don't think we're allowed to group results from disparate studies unless they are all following some common procedure. $\endgroup$
    – daniel
    Jun 12, 2014 at 1:00

1 Answer 1


For some statistical models, see e.g. The Mathematics of Biodiversity (Part 1) by John Baez.

However, in terms of new species, such statistical methods may not suffice - usually also our tools for discovering change. So it might be more accurate to extrapolate number of some kind species we know as a function of time.

For extrapolations (though not for viruses) see e.g.:

In any case, for viruses (and many other fast-multiplying asexual creatures) defining number of species can be anywhere from tricky to impossible (for fundamental reasons, not only - practical). At best you can try to estimate their time to the last common ancestor. See also links in this discussion on Azimuth.

  • $\begingroup$ @daniel Fixed, thanks (sometimes I have issues with copy-paste). BTW: AFAIK OP means 'Original Poster' (i.e. you, in this context), not 'Original Post'. $\endgroup$ Jun 13, 2014 at 13:11
  • $\begingroup$ @daniel I feel overnerded by you (and thanks, it is always good to know). In any case, in the context of Stack Exchange (unlike fora, separating Q from A), I see people using OP when they mean Original Poster (as it is easy to refer to question as 'the question', or Q). $\endgroup$ Jun 13, 2014 at 13:44
  • $\begingroup$ I was thinking about 'outnerded' (it sounds right), but decided to go for 'overnerded' (compare: 'outpowered' vs 'overpowered'). But you are an expert on words here. $\endgroup$ Jun 13, 2014 at 13:55
  • $\begingroup$ OK, here goes. Using MathSE because it's bigger--questions 8623,513239,774176 Poster; questions 708644,272591,350485 Post. I think most refer to poster because it's a service-oriented group. But... $\endgroup$
    – daniel
    Jun 13, 2014 at 14:00
  • $\begingroup$ Let us continue this discussion in chat. $\endgroup$ Jun 13, 2014 at 14:16

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .