Math AI: Modelling with Sinusoidal Functions

From the predictable rise and fall of ocean tides to the daily oscillation of your body temperature, our world is full of repeating patterns. Modelling these cycles mathematically allows you to predict future behavior, analyze trends, and make informed decisions. In IB Mathematics AI, you master this by fitting sinusoidal functions—sine and cosine waves—to real-world periodic data, transforming observed rhythms into powerful predictive equations.

The Anatomy of a Sinusoidal Model

A general sinusoidal function can be written in two equivalent forms: $$y=a\sin (b(x-c))+d$$ or $$y=a\cos (b(x-c))+d$$

Each parameter controls a specific feature of the wave, and your first task is to interpret them from context or data. The amplitude, represented by $|a|$, is the vertical stretch. It measures half the distance between the maximum and minimum values, telling you the intensity of the oscillation—for example, the height of a tide above or below the average sea level.

The period determines the horizontal stretch and is calculated as $T=\frac{2\pi }{|b|}$. This is the length of one complete cycle, such as 24 hours for a daily temperature pattern or roughly 12.4 hours for a semi-diurnal tide. The vertical shift, $d$, is the central equilibrium value, often the average of the maximum and minimum data points. Finally, the phase shift, $c$, indicates the horizontal translation. For a cosine model, $c$ is often the $x$-coordinate of a maximum point; for a sine model, it's often the $x$-coordinate where the curve crosses the midline while increasing.

From Data to Parameters: A Step-by-Step Process

Before using technology, you should be able to estimate model parameters directly from a data set or a description. This builds intuition and provides a good initial check for your regression results. Follow this logical sequence:

1. Find the Vertical Shift ($d$): Calculate the midline: $d=\frac{\text{(Maximum Value)}+\text{(Minimum Value)}}{2}$. In a model for average monthly temperature, $d$ represents the annual average temperature.
2. Find the Amplitude ($a$): Calculate: $a=\frac{\text{(Maximum Value)}-\text{(Minimum Value)}}{2}$. This represents the variation from the average.
3. Find the Period and thus $b$: Measure the horizontal distance between two consecutive peaks (or troughs) to get the period $T$. Then solve for $b$ using $b=\frac{2\pi }{T}$.
4. Find the Phase Shift ($c$): This step requires choosing between a sine or cosine model. Identify a key feature in the data. If you use a cosine model and you have the $x$-coordinate of a maximum point, then $c$ equals that $x$-coordinate. If you use a sine model and have the $x$-coordinate where the data crosses the midline while increasing, then $c$ equals that $x$-coordinate.

For example, consider a simplified tide with a maximum of 8 meters at 3 AM and a minimum of 2 meters 6 hours later. The vertical shift is $d=(8+2)/2=5$ meters. The amplitude is $a=(8-2)/2=3$ meters. The time between a max and the next min is half a period, so the full period $T=12$ hours, giving $b=2\pi /12=\pi /6$. Using a cosine model that starts at a max, the phase shift is $c=3$ (hours after midnight). The model is: $h(t)=3\cos \left(\frac{\pi }{6}(t-3)\right)+5$, where $h$ is height in meters and $t$ is time in hours.

Leveraging Your GDC for Sinusoidal Regression

While the manual method is instructive, real data is messy. Your Graphical Display Calculator (GDC) is essential for finding the optimal sinusoidal model through sinusoidal regression. The process involves entering your data into lists, plotting it to visually confirm periodicity, and then executing the regression command.

On most GDCs (like TI-Nspire or Casio fx-CG50), you will find a "Sine Regression" or "Periodic Regression" tool. You input your $x$ and $y$ data lists. The calculator then uses an iterative algorithm to find the parameters $a$, $b$, $c$, and $d$ that minimize the vertical distances between the data points and the sine curve—a method known as least-squares fitting. It is crucial to provide a good initial guess for the period, which the calculator may request, to help the algorithm converge to the correct solution. The output will be a complete equation in the form $y=a\sin (bx+c)+d$ or similar, which you can graph over your data to assess the fit visually.

Evaluating Your Model: Beyond the Curve

A curve that looks good visually isn't enough; you must statistically evaluate the goodness of fit of your sinusoidal model. Your primary tool for this is the coefficient of determination, denoted $R^2$. When you perform a regression on your GDC, this value is typically provided alongside the parameters.

$R^2$ measures the proportion of the variation in the $y$-data that is explained by the variation in $x$ through your model. Its value ranges from 0 to 1 (or 0% to 100%). An $R^2$ value of 0.93, for instance, means 93% of the change in the dependent variable can be predicted by the sinusoidal model you've fitted. A high $R^2$ (e.g., >0.9) generally indicates a strong fit for biological or physical periodic data, like circadian rhythm data or seasonal temperature changes. However, you must also apply common sense: does the model make contextual sense? Does it predict reasonable values outside the observed data range? A model for Arctic temperature should not predict temperatures above 30°C, for example.

Common Pitfalls

1. Misidentifying the Period: A frequent error is confusing the time between a maximum and a minimum with the full period. Remember, the period is the time for one full cycle (e.g., peak to peak). If data shows a max at day 10 and a min at day 20, the time difference is only half the period. The full period would be 20 days.

• Correction: Always identify two consecutive points with the same phase (two peaks or two troughs) to measure the period directly.

1. Ignoring the Vertical Shift: Students often force the model's midline to be $y=0$, especially when plotting data that appears to oscillate symmetrically. Most real-world data, from tide levels to sound waves, oscillates around a non-zero average.

• Correction: Always calculate $d$ explicitly using the formula for the midline before attempting to find the amplitude.

1. Forcing a Sine Model When Cosine is More Natural: While sine and cosine are horizontally shiftable versions of each other, choosing the wrong base function can make finding the phase shift unnecessarily complicated.

• Correction: Let the data guide you. If the dataset starts at or near a maximum or minimum, use a cosine model. If it starts near the midline, a sine model may be simpler. Your GDC's regression will handle either, but for manual derivation, choose the path of least resistance.

1. Over-relying on Regression Output: Blindly accepting the GDC's parameters without checking their reasonableness can lead to models that are mathematically optimal but contextually absurd (e.g., a negative amplitude for a tide height model).

• Correction: Use your manually estimated parameters as a sanity check. If the GDC gives an amplitude of 50 for a temperature model you know varies by 15 degrees, you likely entered the data incorrectly or need to provide a better initial guess for the period.

Summary

• Sinusoidal functions of the form $y=a\sin (b(x-c))+d$ or $y=a\cos (b(x-c))+d$ are the primary tools for modelling periodic real-world phenomena like tides, temperature cycles, and biological rhythms.
• The parameters have direct physical interpretations: amplitude ($|a|$) for intensity, period ($\frac{2\pi }{|b|}$) for cycle length, vertical shift ($d$) for the central equilibrium, and phase shift ($c$) for horizontal alignment.
• You can estimate these parameters manually from data by systematically finding the midline, amplitude, period, and a key feature point.
• For accurate fitting with real data, use your GDC's sinusoidal regression tool, providing sensible initial guesses to help the algorithm.
• Always evaluate the goodness of fit using the coefficient of determination ($R^2$) and by critically assessing whether the model's predictions make logical sense within the real-world context.