The Unofficial Google Data Science Blog

JANUARY 31, 2017

We can estimate the causal effect of prominence as the difference in log odds of uptake between stories which were recommended and those which were eligible (i.e. The former approach will tend to produce more realistic outcomes, but can be harder to understand, and may not give us adequate data to assess unlikely decisions.

Machine Learning 73

Machine Learning Modeling Statistics Predictive Modeling 73

The Unofficial Google Data Science Blog

DECEMBER 14, 2021

Some of that uncertainty is the result of statistical inference, i.e., using a finite sample of observations for estimation. This blog post introduces the notions of representational uncertainty and interventional uncertainty to paint a fuller picture of what the practicing data scientist is up against. For instance, fluff.ai

Uncertainty 56

Uncertainty Statistics Measurement Cost-Benefit 56

The Unofficial Google Data Science Blog

MARCH 13, 2017

By MUKUND SUNDARARAJAN, ANKUR TALY, QIQI YAN Editor's note: Causal inference is central to answering questions in science, engineering and business and hence the topic has received particular attention on this blog. In this post, we explore causal inference in this setting via the problem of attribution in deep networks.

IT 68

IT Visualization Modeling Uncertainty 68

The Unofficial Google Data Science Blog

APRIL 17, 2017

For us, demand for forecasts emerged from a determination to better understand business growth and health, more efficiently conduct day-to-day operations, and optimize longer-term resource planning and allocation decisions. Forecasting" for us also did not mean using time series in a causal inference setting. OTexts, 2014.

Forecasting 83

Forecasting Modeling Statistics Uncertainty 83

Domino Data Lab

MAY 8, 2019

IMO, the theme here is to shift more of the deeper aspects of inference – such as explainability, auditability, trust, etc. Instead, consider a “full stack” tracing from the point of data collection all the way out through inference. Finale Doshi-Velez, Been Kim (2017-02-28) ; see also the Domino blog article about TCAV.

Machine Learning 105

Machine Learning Data Science Modeling Visualization 105

The Unofficial Google Data Science Blog

JULY 31, 2016

By JEAN STEINER Randomized A/B experiments are the gold standard for estimating causal effects. This blog post explores how this problem arises in applications at Google, and how we can 'mind our units' through analysis and logging. The plot shows how ignoring the group structure (red) leads to increasingly optimistic inferences (i.e.,

Experimentation 42

Experimentation Testing Measurement Metrics 42

The Unofficial Google Data Science Blog

MARCH 22, 2018

This blog post provides details for how we can make inferences without waiting for complete uptake. Background At Google, experimentation is an invaluable tool for making decisions and inference about new products and features. In any case, it is clear that we cannot make inference without some caution in this case.

Experimentation 62

Experimentation Measurement Modeling Statistics 62

The Unofficial Google Data Science Blog

OCTOBER 21, 2015

by NIALL CARDIN, OMKAR MURALIDHARAN, and AMIR NAJMI When working with complex systems or phenomena, the data scientist must often operate with incomplete and provisional understanding, even as she works to advance the state of knowledge. We’re not going to engage in this debate but in this blog post we do focus on science.

Slice and Dice 45

Slice and Dice Experimentation Data-driven Data Science 45

The Unofficial Google Data Science Blog

OCTOBER 12, 2017

Implicitly, there was a prior belief about some interesting causal mechanism or an underlying hypothesis motivating the collection of the data. Understanding the goals of the organization as well as guiding principles for extracting value from data are both critical for success in this environment. And for good reason! Is it real?

Experimentation 61

Experimentation Testing Statistics Metrics 61