1: \begin{abstract}

2: 

3: We entertain the idea that robust theoretical expectations can become

4: a tool in removing hidden observational or data-reduction

5: biases. We illustrate this

6: approach for a specific problem associated with gravitational

7: microlensing.

8: Using the fact that a group is more than just a collection of

9: individuals, we derive formulae

10: for correcting the distribution of the dimensionless impact parameters of 

11: events, $u_{\rm min}$. We refer to the case when undetected biases in the

12: $u_{\rm min}$-distribution can be alleviated by multiplication of

13: impact parameters of all events by a common 

14: constant factor. We show that in this case the general maximum likelihood 

15: problem of solving an infinite number of equations reduces to two constraints, 

16: and we find an analytic solution. 

17: Under the above assumptions, this solution represents a state in which the 

18: ``entropy'' of a microlensing ensemble is at its maximum, that is, the 

19: distribution of $u_{\rm min}$ resembles a specific, theoretically

20: expected, box-like distribution 

21: to the highest possible extent. We also show that this technique does not 

22: allow one to correct the parameters of individual events on the event by 

23: event basis independently from each other.

24: 

25: Subject Headings: dark matter --- gravitation --- gravitational lensing --- methods: statistical --- stars: fundamental parameters (luminosities, masses)

26: \end{abstract}

27: