Number of EPO applications in Docdb families

I've been recently asked whether the number of epo applications would be a good proxy for docdb patent family. In reality it dipends from the average number of EPO applications in each of those families.
I expected the figure to be around 1 but I made a few calcs and, after removing a number of patent fmailies without any EPO (non PCT) application, I got a 1.0887 that is quite confirming the hypotesys.

Here the full counts of number of docdb family by number of EP applications.

N EP	N docdb fam
0	47930863
1	3002328
2	106275
3	15404
4	4537
5	1738
6	846
7	440
8	240
9	144
10	93
11	74
12	47
13	41
14	22
15	20
16	16
17	9
18	9
19	15
20	6
21	10
22	1
23	3
24	4
25	3
26	4
28	2
29	1
30	1
31	1
36	1
38	1
42	1
44	2
46	1
47	1
58	1
331	1
1650	1
3645	1
99668	1

Crosscitations between patent families

Even if Patstat provides only citation file by publications, lots of scientific literature uses citations across patent families as a more relevant measure for knowledge spillovers, patent value etc. since, compared to sheer publication based citation count, it normalizes the data across patent offices.

Nevertheless two caveat should be taken in count when using patent family based citation measures

1) Cross citation may happen
consider example below: all is fine until we look at publications. When using docdb families a crosscitation arises.

citing	cited	citing fam	cited fam
US7507793	US7534577	26973151	26947173
US7534577	US7402400	26947173	26973151

These cases are about 581779 (0,5% of total citations).

 2) self citations are not uncommon

262781 citations @ 2014b edition are within the same docdb family (0.2 %)

citing	cited	citing fam	cited fam
US7534577	WO02064631	26947173	26947173