Question 42·Hard·Two-Variable Data: Models and Scatterplots
A biologist recorded the size of a bacterial culture every hour. The measurements are shown below.
| Time, (hours) | 0 | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|---|
| Culture size, (in thousands of cells) | 1.0 | 1.6 | 2.5 | 3.9 | 6.2 | 9.8 |
Based on these data, which type of model is most appropriate for describing as a function of ?
For data-model questions, quickly scan the table or scatterplot to identify the pattern of growth: check first differences for a constant amount added (linear), then second differences for a constant change in the differences (quadratic), then ratios for a roughly constant factor (exponential), and finally whether growth slows and levels off (logarithmic). Use this systematic check to rule out model types in under a minute, instead of trying to eyeball the exact formula.
Hints
Look at how much changes each hour
Find the increase in culture size from one hour to the next (for example, from 1.0 to 1.6, from 1.6 to 2.5, etc.). Are these increases constant, getting bigger, or getting smaller?
Use differences to rule out some models
Ask yourself: which type of model would have a constant increase each hour? Which would have increases that change in a regular pattern? Does the table show either of those behaviors?
Try comparing ratios instead of differences
Divide each culture size by the previous one (for example, , , and so on). Are these ratios close to each other, or do they vary a lot? What kind of growth is suggested when values multiply by about the same factor each step?
Desmos Guide
Enter the data as a table
Create a table, put the time values in the first column (as ), and the culture sizes in the second column (as ). This lets you see the scatterplot.
Visually compare to a straight line
Type y1 ~ m x1 + b to perform a linear regression. Look at how well the line follows the plotted points. If the points clearly bend away from the line and the errors grow, a linear model is not ideal.
Try a quadratic and a logarithmic fit
Type y1 ~ a x1^2 + b x1 + c for a quadratic regression and y1 ~ a + b ln(x1) for a logarithmic regression. Compare how closely each curve follows the plotted points across the entire range of values.
Try a multiplicative-growth model and choose the best fit
Type y1 ~ a b^{x1} to fit a model where values grow by a roughly constant factor each step. Compare the curves from all your regressions and see which one hugs the data points most closely over all times; the type of that best-fitting model is the answer choice you should select.
Step-by-step Explanation
Check whether a linear model fits (constant differences)
Look at how much the culture size increases each hour.
From the table:
- From to :
- From to :
- From to :
- From to :
- From to :
These differences are not constant; they are getting larger, so a simple increasing linear model (which needs constant differences) does not fit well.
Check whether a quadratic model fits (pattern in the differences)
For a quadratic model, the second differences (the differences of the differences) are roughly constant.
First differences (already found): .
Second differences:
These second differences are not close to constant, so a quadratic model is not a good match either.
Think about the type of growth: additive or multiplicative
When values grow by adding about the same amount each step, that suggests one kind of model; when they grow by multiplying by about the same factor each step, that suggests a different model.
Compute the ratios of consecutive culture sizes:
These ratios are all around , meaning each hour the culture is multiplying by nearly the same factor, not just adding a fixed amount.
Compare with the remaining model types and conclude
An increasing logarithmic model would have growth that slows down over time (the differences between terms would get smaller), but here the differences are getting larger and the growth speeds up. The data show values that increase faster and faster, with a nearly constant ratio between consecutive terms, which is characteristic of increasing exponential growth.
So the most appropriate model for as a function of is increasing exponential (choice C).