Databeers Dub #1 - Cathal Walsh - Statistics in the Big Data World

Cathal WalshChair in Statistics

HRB Research Leader Health Decision Science

Model CalibrationIdentify Utility of Decision Makers

Evidence Synthesis

BIGDATA

?

Real Decisions!

“Just because they haven’t seen it there, doesn’t mean it isn’t

present.”

BIGDATA =

= small data

Model FreeN=“All”

N=“All”?Others, however, especially pattern detection problems …, cannot be based on a sample. Moreover, in some cases one has the entire data set available (in the numerical examples above, all of the Wal-Mart transactions for a year will be available, not merely a sample of them).

(Hand et al, 2000.)

Who did the counting? What did they count, and why? How did they go about it?

Statistics as construction (Best 2005)

Model Free? Human Decisions:

Probability of Death

S(T=5 | M,X, Diet) vs S(T=5 | M,X, No Diet)

Boiled Model Syndrome

Glimcher (2004)

Bias? (Meng 2016)

Variability of Measure

Sample Size

Random Sampling?

I = indicator variable for Sample

Framework revisited:

Science / Industry

Computing

Maths / Stats

Prediction? Classification … Health Analytics