S3E06: Item Parcels: What Could Possibly Go Wrong?!
Quantitude
00:00
How to Model for 4,000 Bivariate Correlations
The idea of modeling for 1000 correlations is incredibly unwieldy. So being able to reduce this down to something smaller feels like a good thing to do. It also has the potential to get around some problems. For example, the distribution of scores that we might have on any particular item might be such that if the item is continuous, perhaps the responses are skewed.
Transcript
Play full episode
