Estimating Derivatives of a Function at a Point

A derivative is a slope, and slope just needs two points. When you have a formula, you can take a limit all the way down to a single point. When you only have a table of measured values, you can't take that limit, so you approximate it with a nearby slope. Pick a point before a and a point after a, find the slope between them, and that slope estimates f'(a). This works because f'(a) really is the limit of slopes between a and nearby points, so a slope between two close points that straddle a is a reasonable stand-in.

The symmetric difference quotient uses one point before a and one point after, equally spaced: [f(a+h) - f(a-h)] / (2h). Compare this to a one-sided quotient like [f(a+h) - f(a)] / h, which only sees what happens on one side. A curve can bend, so the slope just to the left of a point and just to the right can genuinely differ. The symmetric version averages information from both directions, which cancels out some of that bending and gets closer to the true instantaneous slope. That's also why it ignores the middle value f(a) entirely: it only needs the two outer points.

This connects directly to the formal limit definition of the derivative that comes right before it in the unit. The symmetric quotient is the same idea, just stopped at a finite step size h instead of pushed to the limit as h approaches 0. Later, when rules like the power rule and product rule take over for computing exact derivatives from formulas, this table-based estimation remains the tool for situations with only numerical data, like a speedometer reading recorded every few seconds.