2

To test the hypothesis that parents name their children without bias towards their own initials we have data of one parent name and one of their children's names. There are 600 data points and it has been measured that roughly 7% of parents have names that start with the same first initial as their own.

If there is no such tendency, what is the expected rate of parent-child first initial matches? Is it as simple as 1 in 26 (letters in the alphabet)? What would be the best way to go about it? Could you use distribution of children's names?

CallumDA
  • 173
  • You should use the distribution of children's names and parents names. E.g. you can consider the parents name mostly started with (say >50%) C as in Claus, Candy,... while todays children's names start mostly with J. Hence you should compare the (different) distributions the initials for parents and for children have. It is obviously not as simple as 1 in 26, since very few people have initial y or x. (there might also be a correlation the other way, like people called Kevin are more common in a certain subset of people, where later on the name Jaqueline is common) – ctst Sep 06 '16 at 13:07
  • 1
    The null hypothesis must be that all parents name their children randomly with names pulled from the same distribution -- but it will be strictly counterfactual to assume that all initials are equally likely in such a distribution. As an extreme example, if we're in a culture where half of the given names start with K, then the expected frequency of same-initial cases would be at least 25% just because of that, even without any parental bias involved. – hmakholm left over Monica Sep 06 '16 at 13:07
  • The appropriate distributions for children’s and parents’ names will also depend on the source of your data. If the data are English, you might want to look here for information on actual naming practices; if U.S., try here. – Brian M. Scott Sep 06 '16 at 17:42
  • 1
    Can't be as simple as counting 26 letters. Not all letters A-Z are equally likely for names. – BruceET Sep 06 '16 at 22:30
  • With 600 data points, you have distributions of parents' and childrens' initials in the data. So you can find the expected number you would expect to match from that, and see whether the actual number of matches related to that. – Henry Sep 07 '16 at 08:01

0 Answers0