feat[matching]: Mendez method for imputing missing answers

[kaljarv]

Mendez (2017, 50–51) suggests using recommendation scores that range from -100 to +100. This operationalization of the matching score has the benefit that missing party/candidate responses can be incorporated to the matching, so that they make the matching scores gravitate towards the middle point of the scale (zero). Mendez (2017) suggests that one divides the sum of matching scores with the candidate by the number of statements to which the user has responded. For example, if the user has all 30 statements, but the candidate has responded to only 15 statements, and in those 15 statements they have responded exactly the same way (sum of matching scores: 15/15) the agreement would be 15/30 = +50 in the range of -100 to +100.