Differential Approximation: Using \(dy = f'(x)dx\) to estimate the change in function value due to a small change in input.
- Identify: The function \(f(x)\) and base point \(x\)
- Calculate: The derivative \(f'(x)\)
- Determine: The change in input \(dx = \Delta x\)
- Apply: The formula \(dy = f'(x)dx\) to estimate \(\Delta y\)
Area function: \(A(r) = \pi r^2\)
Base radius: \(r = 10\) cm
Measurement error: \(dr = 0.1\) cm
Formula: \(dA = A'(r)dr\)
\(A(r) = \pi r^2\)
\(A'(r) = 2\pi r\)
\(dA = A'(r)dr = 2\pi r \cdot dr\)
\(dA = 2\pi (10)(0.1) = 2\pi = 6.28\) cm²
Original area: \(A = \pi (10)^2 = 100\pi \approx 314.16\) cm²
Relative error: \(\frac{6.28}{314.16} \approx 0.02 = 2\%\)
Maximum error in area: 6.28 cm²
Relative error: 2%
The area is sensitive to radius measurement errors!
• Power Rule: \(\frac{d}{dr}[r^n] = nr^{n-1}\)
• Error Propagation: \(dA = A'(r)dr\)
• Sensitivity: Area is twice as sensitive as radius
Volume function: \(V(r) = \frac{4}{3}\pi r^3\)
Base radius: \(r = 5\) cm
Measurement error: \(dr = 0.05\) cm
\(V(r) = \frac{4}{3}\pi r^3\)
\(V'(r) = 4\pi r^2\)
\(dV = V'(r)dr = 4\pi r^2 \cdot dr\)
\(dV = 4\pi (5)^2(0.05) = 4\pi (25)(0.05) = 5\pi \approx 15.71\) cm³
Original volume: \(V = \frac{4}{3}\pi (5)^3 = \frac{500\pi}{3} \approx 523.60\) cm³
Relative error: \(\frac{15.71}{523.60} \approx 0.03 = 3\%\)
Maximum error in volume: 15.71 cm³
Relative error: 3%
Volume is even more sensitive to radius errors than area!
• Power Rule: \(\frac{d}{dr}[r^3] = 3r^2\)
• Constant Multiple: \(\frac{d}{dr}[cf(r)] = c \cdot f'(r)\)
• Higher Sensitivity: Cubic functions are more sensitive than quadratic
Compound interest function: \(A(r) = P(1+r)^t\)
Parameters: \(P = 1000\), \(t = 10\), \(r = 0.05\)
Rate error: \(dr = 0.001\)
\(A(r) = 1000(1+r)^{10}\)
\(A'(r) = 1000 \cdot 10(1+r)^9 = 10000(1+r)^9\)
\(dA = A'(r)dr = 10000(1+r)^9 \cdot dr\)
\(dA = 10000(1.05)^9(0.001) = 10(1.05)^9\)
\(dA = 10 \times 1.5513 \approx 15.51\) dollars
Original amount: \(A = 1000(1.05)^{10} \approx 1628.89\) dollars
Relative error: \(\frac{15.51}{1628.89} \approx 0.0095 = 0.95\%\)
Maximum error in final amount: $15.51
Relative error: 0.95%
Even small interest rate errors can affect long-term investments!
• Chain Rule: \(\frac{d}{dr}[(1+r)^t] = t(1+r)^{t-1}\)
• Exponential Growth: Small errors compound over time
• Financial Sensitivity: Long-term investments are sensitive to rate changes
Differential: \(dx\) and \(dy\) represent infinitesimal changes in \(x\) and \(y\)
Increment: \(\Delta x\) and \(\Delta y\) represent finite changes in \(x\) and \(y\)
Error Estimation: Quantifying uncertainty in computed quantities
Absolute error: \(|dy|\) estimates \(|\Delta y|\)
Relative error: \(\frac{|dy|}{|f(x)|}\) indicates percentage uncertainty
• Continuity: \(f\) and \(f'\) must be continuous near the base point
• Small change: \(|dx|\) should be much smaller than \(|x|\)
• Smooth behavior: Function should not have rapid oscillations
• First-order approximation: Higher-order terms should be negligible
Area function: \(A(l,w) = lw\)
Dimensions: \(l = 10\) cm, \(w = 15\) cm
Errors: \(dl = 0.1\) cm, \(dw = 0.1\) cm
\(A(l,w) = lw\)
\(\frac{\partial A}{\partial l} = w = 15\)
\(\frac{\partial A}{\partial w} = l = 10\)
\(dA = \frac{\partial A}{\partial l}dl + \frac{\partial A}{\partial w}dw = w \cdot dl + l \cdot dw\)
\(dA = 15(0.1) + 10(0.1) = 1.5 + 1.0 = 2.5\) cm²
Original area: \(A = 10 \times 15 = 150\) cm²
Relative error: \(\frac{2.5}{150} \approx 0.0167 = 1.67\%\)
Maximum error in area: 2.5 cm²
Relative error: 1.67%
Both dimension errors contribute to the total area error!
• Partial Derivatives: \(\frac{\partial}{\partial l}[lw] = w\), \(\frac{\partial}{\partial w}[lw] = l\)
• Total Differential: Sum of individual contributions
• Error Addition: Independent errors add linearly
Period function: \(T(L,g) = 2\pi\sqrt{\frac{L}{g}}\)
Parameters: \(L = 1\) m, \(g = 9.8\) m/s²
Errors: \(dL = 0.01\) m, \(dg = 0.1\) m/s²
\(T(L,g) = 2\pi\sqrt{\frac{L}{g}}\)
\(\frac{\partial T}{\partial L} = 2\pi \cdot \frac{1}{2\sqrt{\frac{L}{g}}} \cdot \frac{1}{g} = \frac{\pi}{\sqrt{Lg}}\)
\(\frac{\partial T}{\partial g} = 2\pi \cdot \frac{1}{2\sqrt{\frac{L}{g}}} \cdot \left(-\frac{L}{g^2}\right) = -\frac{\pi\sqrt{L}}{g^{3/2}}\)
\(dT = \frac{\partial T}{\partial L}dL + \frac{\partial T}{\partial g}dg\)
\(\frac{\partial T}{\partial L} = \frac{\pi}{\sqrt{1 \cdot 9.8}} = \frac{\pi}{\sqrt{9.8}} \approx \frac{3.1416}{3.1305} \approx 1.003\)
\(\frac{\partial T}{\partial g} = -\frac{\pi\sqrt{1}}{9.8^{3/2}} = -\frac{\pi}{9.8\sqrt{9.8}} \approx -\frac{3.1416}{30.68} \approx -0.102\)
\(dT = 1.003(0.01) + (-0.102)(0.1) = 0.01003 - 0.0102 = -0.00017\)
Maximum error: \(|dT| = 0.00017\) seconds
Maximum error in period: 0.00017 seconds
The gravitational acceleration error nearly cancels the length error!
This demonstrates how different parameter errors can interact.
• Chain Rule: For composite functions, use partial derivatives
• Error Cancellation: Opposite-signed errors can partially cancel
• Dimensional Analysis: Check units of each error contribution
Differential Approximation: The process of estimating the change in a function's output due to small changes in its inputs using the derivative.
Total Differential: For functions of multiple variables, the sum of all partial derivatives times their respective input changes.
Error Propagation: The process of determining how uncertainties in input variables affect the uncertainty in the output.
- Identify the function: Determine \(f(x)\) or \(f(x_1, x_2, ..., x_n)\) and all relevant variables
- Calculate derivatives: Find \(f'(x)\) or all partial derivatives \(\frac{\partial f}{\partial x_i}\)
- Determine input errors: Identify all possible errors \(dx_i\) in the input variables
- Apply differential formula: Use \(df = \sum \frac{\partial f}{\partial x_i}dx_i\) for multiple variables
- Analyze results: Calculate absolute and relative errors, interpret significance
• Single variable: \(dy = f'(x)dx\)
• Multiple variables: \(df = \sum_{i=1}^{n} \frac{\partial f}{\partial x_i}dx_i\)
• Relative error: \(\frac{df}{f} = \frac{f'(x)}{f(x)}dx\)
• Error bound: \(E = \frac{M}{2}(dx)^2\) where \(M = \max|f''(t)|\)
- Linear: \(f(x) = x\) (error amplification factor = 1)
- Quadratic: \(f(x) = x^2\) (error amplification factor = 2x)
- Exponential: \(f(x) = e^x\) (error amplification factor = e^x)
Sensitivity analysis: Different functions have different sensitivities to input errors.
- Linear functions have constant sensitivity
- Quadratic functions become more sensitive as x increases
- Exponential functions become extremely sensitive quickly