Sensitivity (Meta)

Meta
principle

Also known as: sensitivity analysis, ill-conditioning

Grade 9-12

View on concept map

The degree to which a result or output changes in response to small changes in its inputs, parameters, or assumptions. High sensitivity means small errors in inputs cause large errors in outputs — knowing this guides where to spend effort on precision in a calculation.

Definition

The degree to which a result or output changes in response to small changes in its inputs, parameters, or assumptions.

💡 Intuition

Is this result stable, or does a tiny change blow everything up?

🎯 Core Idea

High sensitivity means results are unreliable or require high precision.

Example

Quadratic formula: small change in coefficients \to small change in roots (usually). But near discriminant = 0, very sensitive.

Formula

\text{sensitivity} \approx \frac{\Delta\text{output}}{\Delta\text{input}} (how much the output changes per unit change in input)

Notation

\Delta denotes a small change; high \frac{\Delta\text{output}}{\Delta\text{input}} means high sensitivity

🌟 Why It Matters

High sensitivity means small errors in inputs cause large errors in outputs — knowing this guides where to spend effort on precision in a calculation.

💭 Hint When Stuck

Compute the answer with the original inputs, then recompute with a slightly changed input (say, add 0.01). Compare the two outputs; a large difference signals high sensitivity.

Formal View

The condition number \kappa = \left|\frac{x \cdot f'(x)}{f(x)}\right| measures relative sensitivity; \kappa \gg 1 means the problem is ill-conditioned

🚧 Common Stuck Point

Sensitivity is not the same as magnitude — a model can produce large outputs while being insensitive to inputs, or produce small outputs while being highly sensitive.

⚠️ Common Mistakes

  • Ignoring sensitivity and trusting a computed answer blindly — near a sensitive region, small rounding errors can produce wildly wrong results
  • Not recognizing when a problem is ill-conditioned — e.g., solving nearly singular linear systems gives unreliable answers
  • Confusing sensitivity of the problem with sensitivity of the method — even a good algorithm fails on an inherently ill-conditioned problem

Frequently Asked Questions

What is Sensitivity (Meta) in Math?

The degree to which a result or output changes in response to small changes in its inputs, parameters, or assumptions.

What is the Sensitivity (Meta) formula?

\text{sensitivity} \approx \frac{\Delta\text{output}}{\Delta\text{input}} (how much the output changes per unit change in input)

When do you use Sensitivity (Meta)?

Compute the answer with the original inputs, then recompute with a slightly changed input (say, add 0.01). Compare the two outputs; a large difference signals high sensitivity.

Next Steps

robustness

How Sensitivity (Meta) Connects to Other Ideas

To understand sensitivity (meta), you should first be comfortable with local vs global behavior. Once you have a solid grasp of sensitivity (meta), you can move on to robustness.

← Back to Explorer