Free cheat sheets

Precalculus / Chapter 5

Exponential Functions

Cheat sheet

Exponential Functions

  • An exponential f(x)=Abxf(x)=Ab^x has b>0b>0, b≠1b\ne1 and A≠0A\ne0. For equal input increments h, f(x+h)/f(x)=bhf(x+h)/f(x)=b^h. With A>0A>0, b>1b>1 grows and 0<b<10<b<1 decays.
  • For f(x)=80(0.75)xf(x)=80(0.75)^x, each unit multiplies the output by 0.75, a 25% decrease. A negative A reflects the graph; multiplication factors remain the same, but increasing/decreasing language must account for the sign.
  • The unshifted graph has y-intercept A and horizontal asymptote y=0y=0. Positive A gives positive outputs; it never reaches zero at a finite input.

Exponential Function Manipulation

  • For positive bases, bx+y=bxbyb^{x+y}=b^xb^y, bx−y=bx/byb^{x-y}=b^x/b^y, and (bx)r=bxr(b^x)^r=b^{xr}. Changing the time unit changes the exponent, not just the label.
  • If a value multiplies by 1.21 every two years, its model is A(1.21)t/2=A(1.1)tA(1.21)^{t/2}=A(1.1)^t. For Abx−h+kA b^{x-h}+k, the horizontal asymptote is y=ky=k; if A is positive, the range is (k,∞)(k,\infty).

Exponential Function Context and Data Modeling

  • For positive data and y=Abxy=Ab^x, b=(y2/y1)1/(x2−x1)b=(y_2/y_1)^{1/(x_2-x_1)} and A=y1/bx1A=y_1/b^{x_1}. Values 10 at x=1x=1 and 40 at x=3x=3 give b=2b=2, A=5A=5.
  • Doubling time d gives y=y02t/dy=y_0 2^{t/d}; half-life h gives y=y0(1/2)t/hy=y_0(1/2)^{t/h}. The elapsed-time unit must match the period’s unit.
  • Continuous change uses y=y0ekty=y_0e^{kt}; k is a continuous rate, while ek−1e^k-1 is the fractional change per unit time. For k=ln⁡2/3k=\ln2/3, doubling takes 3 time units. Do not extend a growth model indefinitely without checking its assumptions.

Competing Function Model Validation

  • With equally spaced inputs, compare first differences (linear), second differences (quadratic), and ratios (exponential). Near-constant values suggest a candidate family; noisy finite data cannot prove it.
  • A residual is e=y−y^e=y-\hat y: observed minus predicted. If a model predicts 12 and the measurement is 10, e=10−12=−2e=10-12=-2. A curved pattern in residuals signals missed structure; small residuals alone do not establish a causal mechanism.
  • Interpolation predicts within the measured input range; extrapolation predicts beyond it and needs stronger justification. Compare models in the actual output units as well as on any transformed scale.

Sign in