When the data’s not there, expert-led models could help

Missing data is a problem. Expert elicitation taps the knowledge of many, say consultants

If the data’s spotty, fleeting, reflective of nothing real or just not there, how does a quant fill the void? The expert elicitation approach to modelling could offer a way.

In contrast to plain vanilla expert judgement, expert elicitation extracts probabilistic belief statements, often from several authorities, on quantities or parameters. The approach brings a structured procedure to building models from sparse datasets, and adheres to the same scientific rules: transparency, accountability