Mark Haggard gave a great talk here today about relating (cheap) subjective and (expensive) objective measures when the subjective measure is guaranteed to be biased. The general gist, I got, was that you calibrate someone’s subjective bias by regressing the subjective measure on the objective measure for a large population. You can then examine the residuals to work out how biased an individual is, and use that to adjust for bias in subsequent measurements. The actual measurement model was a tad more complicated than this.
He said a few things that made me happy.
- He took the time to explain that he didn’t intend to use the word “bias” in an pejorative way. Many biases, he argued, have adaptive value. This is good, as part of my PhD is on interpreting biases in reasoning as being adaptively useful and examining how they relate to each other and to sociocommunicative traits.
- When discussing p-values he mentioned how he didn’t want to get into a silly game of giving exact p’s. He had a huge sample size, all the p’s were tiny. He explained that effect size is what matters. I guess everyone should know this by now, but nice to get it reinforced.
- He argued that it’s wrong to use factor analysis to explore the structure of the mind, rather, he argued, it should be used for data reduction. I need to think a little more about this…