A gardener records the height of a bamboo plant. At week \(0\), the plant is \(50\,\text{cm}\) tall. At week \(2\), it is \(90\,\text{cm}\) tall, and at week \(4\), it is \(170\,\text{cm}\) tall.
a) Find a linear model \(h_{\text{lin}}(t)\) using the data from weeks \(0\) and \(4\). Then find a quadratic model \(h_{\text{quad}}(t)\) using all three data points.
b) At week \(5\), the measured height is \(222.5\,\text{cm}\). Determine which model gives the better prediction and justify your answer.
Hints
- How many points determine a linear model? How many determine a quadratic model?
- Substitute the data points into each model's general form.
- Compare each week-\(5\) prediction with the measured height using absolute error.
Solution
1. For the linear model, \(h_{\text{lin}}(0) = 50\), so the y-intercept is \(50\). The slope through \((0, 50)\) and \((4, 170)\) is \(\frac{170 - 50}{4 - 0} = 30\). Thus, \(h_{\text{lin}}(t) = 30t + 50\).
2. Let \(h_{\text{quad}}(t) = at^2 + bt + c\). From \(h_{\text{quad}}(0) = 50\), \(c = 50\).
3. Using \((2, 90)\) gives \(4a + 2b = 40\), or \(2a + b = 20\). Using \((4, 170)\) gives \(16a + 4b = 120\), or \(4a + b = 30\).
4. Subtract the equations: \(2a = 10\), so \(a = 5\). Then \(2 \cdot 5 + b = 20\), so \(b = 10\). Therefore, \(h_{\text{quad}}(t) = 5t^2 + 10t + 50\).
5. At week \(5\), the linear model predicts \(h_{\text{lin}}(5) = 200\,\text{cm}\), with an absolute error of \(22.5\,\text{cm}\). The quadratic model predicts \(h_{\text{quad}}(5) = 225\,\text{cm}\), with an absolute error of \(2.5\,\text{cm}\). The quadratic model gives the better prediction.
Answer
a) \(h_{\text{lin}}(t) = 30t + 50\)
\(h_{\text{quad}}(t) = 5t^2 + 10t + 50\)
b) The quadratic model is better. Its prediction is \(225\,\text{cm}\), which is \(2.5\,\text{cm}\) from the measured value; the linear prediction is \(200\,\text{cm}\), which is \(22.5\,\text{cm}\) away.