Update README.md

README.md

@@ -73,9 +73,11 @@ Swensson, and Wretman 1992; Haas et al, 1995)
 
 
 ## Complexities
-Edge Case : Where all values seen are unique (d unique values in sequence of length d), no statistical method works and the methods fall back to a special case of [birthday problem](https://en.wikipedia.org/wiki/Birthday_problem) with no collisions. In this problem, we try different values of the distinct values in the population (D), and estimate the probability that we draw d unique values from it with no collision. Intuitively, if our sample contains 10 unique values, then D is more likely to be 100 than 10. If we set a posterior probability (default 0.1), we can then compute the smallest value for D where the probability is greater than 0.1. You can tweak the probability of the birthday solution to get the lower bound (around 0.1) or an upper bound estimate (something like 0.9) of D.
+### All values are unique
+Where all values seen are unique (d unique values in sequence of length d), no statistical method works and the methods fall back to a special case of [birthday problem](https://en.wikipedia.org/wiki/Birthday_problem) with no collisions. In this problem, we try different values of the distinct values in the population (D), and estimate the probability that we draw d unique values from it with no collision. Intuitively, if our sample contains 10 unique values, then D is more likely to be 100 than 10. If we set a posterior probability (default 0.1), we can then compute the smallest value for D where the probability is greater than 0.1. You can tweak the probability of the birthday solution to get the lower bound (around 0.1) or an upper bound estimate (something like 0.9) of D.
 
-Computation of N (population size):
+### Knowledge of population size (N)
+In most real world problems, the population size N will not be known - all that is available is the sample sequence. Most of estimators would be improved if the population size N is given to it, but if it isn't the estimators would just assume a very large N and attempt to estimate D anyway.
 
 ## Additional planned work