Materials and methodsThe estimated population of each locality: The population of all 14 localities as given in the Government Gazette of 29 April 1997, and which refers to the estimated population by locality as of 31st March 1997 is used in these calculations. The total population by this estimate was 31,354 persons. The distribution by locality is also obtained from this data, and is referred to below as "total village/town population" The electoral register: This consists of all those persons registered to vote (i.e. includes those over 18 years of age). It contains all those living in Gozo or having a Gozo I.D. number. The register contained 24,777 persons. However, it was decided to exclude those persons bearing Gozo ID but residing elsewhere. This reduced the number of surnames available for analysis to 22,505 (or 71.8 % of the estimated Gozo population). A number of analyses were performed, using the following definitions and parameters: Frequency (of a particular surname): number of persons having a particular surname. Total frequency: frequency of a particular surname in Gozo Locality Frequency: frequency of a particular surname in the specified locality. Relative frequency: (RF): This is assessed as the number of persons with a particular surname in a locality divided by the total number persons with that surname in Gozo (per cent). A high relative frequency means that a particular surname is found more commonly in this locality than elsewhere. To bear in mind that uncommon surnames may have a very high relative frequency in a particular village. The Observed - Expected frequency (per cent) (O-E%): This is defined as: Observed - Expected number of persons with a particular surname divided by the observed number and multiplied by 100. The expected number is derived from the total number of persons with a particular surname in a locality divided by the total number of persons in Gozo and multiplied by the total number of persons in the locality. A surname having a high RF or O-E% indicates that that particular surname is found more commonly in a particular locality than is expected by chance. Low Relative frequency: defined as those surnames with an O-E% of less than 50. Only the relatively common surnames (with a frequency of at least 20 are considered here). Unbiased Random Isonomy (I) ( see Rodriguez-Larralde et al, 1993): I = Sum(pk)2 / N where p = relative frequency of surname k in each locality N is the number of surnames in the locality. Fisher's alpha is estimated from the formula (Barrai et al 1992): I = 1/alpha
    
| 
.