Exponential data can look curved on ordinary axes but straight on a semi-log plot. The transformation does not change the data; it changes how vertical distances represent values. This makes multiplicative patterns easier to recognize and fit.
Equal distances can represent equal ratios
On an ordinary vertical axis, equal distances represent equal differences, such as 10, 20, 30, 40. On a logarithmic vertical axis, equal distances represent equal ratios, such as 1, 10, 100, 1,000.
A semi-log plot uses a logarithmic scale on one axis, usually , and an ordinary scale on the other. If the vertical coordinate is then moving up one vertical unit means multiplying the original y-value by .
Only positive y-values can be plotted with a real logarithmic transformation. Zero and negative observations require a different model or transformation; they cannot simply be inserted into a log formula.
Why exponential growth becomes a line
Start with where A and are positive and Taking natural logarithms gives
This has linear form if Its intercept is and slope is Therefore and
The slope is a change in log output per input unit. It is not the original additive increase in . Positive slope means multiplicative growth, negative slope means multiplicative decay.
Example: recognize a constant ratio
At times 0, 1, 2, and 3, a quantity has values 5, 10, 20, and 40. The ratios are all 2, suggesting
On a base plot, the transformed outputs are and Their differences are constant at 1, so the points form a line.
On a natural-log plot, the slope would be not 1. The straightness is unchanged, but the numerical slope depends on the logarithm base. Always identify the axis or transformation used.
Taking base-2 logarithms turns each doubling into a vertical increase of one, producing a straight line.
Taking base logarithms turns each doubling into a vertical increase of one, producing a straight line.
Describe this figure
Taking base logarithms turns each doubling into a vertical increase of one, producing a straight line.
Marked points: filled marker at (0, 2.3219); filled marker at (1, 3.3219); filled marker at (2, 4.3219); filled marker at (3, 5.3219).
Example: recover an exponential equation
A fitted line for transformed data is Exponentiating both sides gives
Thus the initial factor is approximately 3.32 and the multiplier per input unit is approximately 1.35. The modeled increase is about 35% per unit, not 30%. The coefficient 0.3 is the natural-log growth rate; the actual multiplier is
For equal to 2, use Do not substitute into the transformed line and report 1.8 as the original y-value; 1.8 is its logarithm.
Decay and doubling time
If with then describes decay. Each increase of one input unit multiplies by
For growth with log-slope positive, doubling requires so With equal to 0.3, the doubling time is about 2.31 input units. This derives a meaningful time scale directly from the transformed slope.
A semi-log straight line indicates an exponential relationship, while a straight line on plot indicates a power relationship. Do not confuse the two plots: transforming both axes changes the model family being linearized.
Assess the fit in the relevant scale
Fitting a line to log-transformed data minimizes a criterion in log units, emphasizing relative discrepancies rather than the same absolute discrepancies in original units. Transforming back does not automatically give the same fit as directly minimizing squared errors in y.
Inspect residual patterns and consider the application. A visually straight transformed plot supports a model choice but does not prove the mechanism or justify unlimited extrapolation.
Try it yourself
A line is Write the corresponding exponential model and find at equal to 10.
Answer: At equal to 10, the log value is 1, so equals 10.
What to remember
A semi-log line represents constant multiplicative change. Read the logarithm base, exponentiate to recover original values, and interpret slope in transformed units.