Physical model

As explained in Causal Inference for Data Science

A physical model is a representation of real-world mechanisms and entities that reproduces system behavior under specified conditions.

A physical model represents how real-world mechanisms produce outcomes under specified conditions. It matters when a decision changes those conditions: a predictive pattern may work for tomorrow, yet fail after an intervention or in an unfamiliar setting. Ask whether the goal is merely to predict what usually happens or to forecast what will happen after the world is changed. section 4.1.3

Physical models are useful when:

• Historical regularities are not enough, such as predicting sunrise on Mars rather than repeating yesterday’s Earth observation. section 4.1.3
• Decisions involve interventions, such as rerouting traffic, closing lanes, or adding hospital resources. section 3.3.3
• Unobserved scenarios must be explored by representing mechanisms and generating outcomes under different conditions. section 3.3.3

Traffic as a simplified physical model

A detailed traffic simulator represents signals, car types, and driver behavior. A causal model can simplify the same problem by learning from existing data and treating traffic policies as interventions. Detectors provide flow, the number of cars passing in an hour, and occupancy, the share of time a detector is covered. Flow alone is ambiguous: low flow may mean light traffic or a traffic jam; occupancy helps distinguish them. section 3.3.3

Figure 3.17 Description of traffic behavior unfolded in time

Use the time-unfolded traffic diagram to connect detector measurements, traffic flow, congestion, and changing system states. section 3.3.3

How Causal Inference for Data Science covers it

How deep it goes

Defines physical models through solar-system and simulation examples, but stops short of detailed simulator construction.

Examples

Uses Mars sunrise prediction, traffic flow and occupancy, rerouting, lane closures, and hospital patient flows.

What it assumes

Assumes familiarity with prediction while contrasting machine learning with causal inference and simulation.

What you can do after

Evaluate whether a prediction task needs a model of mechanisms for conditions unlike historical data.

How it connects

Connects physical models to machine learning, causal inference, interventions, simulators, forecasting, and prediction.

liveBookmanning.com

Read Causal Inference for Data Science on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Hacking CryptographyA physical model relates measurable system behavior to the physical properties of the phenomenon producing it.

Compare how 2 books explain physical model →

Related concepts

• simulatorused by · A simulator is software that computationally models selected behavior of a real or hypothetical system to generate outcomes for testing, analysis, experimentation, or prediction.
• determinismkind of · Determinism is the property of a process or system in which identical inputs and relevant states or conditions determine the same outcome or behavior.
• statistical health testused by · A statistical health test checks whether an entropy source or random-number generator continues to exhibit its expected behavior and entropy properties.
• machine learningcontrasts with · Machine learning is a field of computational methods in which systems learn patterns or relationships from data to make predictions, decisions, or other outputs.
• language modelkind of · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.

This explainer was generated by Manning's knowledge graph from the text of Causal Inference for Data Science. The section links point to the parts of the book it draws on.