and you would be right. Say I have a broken phone which is an Android. A prejudice would be that androids are likely to break. Mathematically we could represent as

  P(broken|android) =~ P(android|broken)

We see from Bayes rule that this is false, since

  P(android) >> P(broken)

I’d argue that prejudice and bias applies most all statistics, but most often we just don’t notice or care. E.g. The biases inherent to specific domain, of those who captured the data, those who selected the data, selection of statistical techniques, real-world outcomes reinforced by usage of these statistics, etc.