Estimating Error in Linear Approximation
What You'll Learn
You'll learn how to quantify the accuracy of linear approximations by estimating the maximum possible deviation between the approximation and the true function value.
The Core Idea
When you use a tangent plane or the total differential to approximate a function's value, you're essentially replacing a curved surface with a flat plane near a point. But how far off might you be?
The error in linear approximation depends on:
- How far you move from the base point – the farther you travel, the worse the approximation
- How much the surface curves – measured by second-order partial derivatives
For a function f(x, y) approximated near point (a, b), the error E is bounded by:
|E| ≤ (1/2)M(|Δx|² + |Δy|²)
where M is the maximum value of all second partial derivatives (|f_xx|, |f_yy|, |f_xy|) in the region you're examining.
Practical Approach
Think of it like this: When you approximate a hillside with a flat board, the error depends on (1) how far you step from your starting point, and (2) how sharply the hill curves. A gently rolling hill gives small errors; a steep valley gives large ones.
Steps to estimate error:
- Find all second-order partial derivatives: f_xx, f_yy, f_xy
- Determine the maximum value M among these in your region of interest
- Measure Δx and Δy (distance from the base point)
- Apply the error bound formula
This gives you a worst-case guarantee: your approximation will be at least this accurate, often much better.
Key Takeaway: The error in linear approximation grows with the square of the distance from the base point and is proportional to the maximum curvature (second derivatives) of the surface.