Overview
This topic runs through every other one. Olympiad questions often describe an experiment you have never seen and ask what it shows. You rarely need special knowledge. You need a careful method.
Use the same routine every time. What was manipulated? What was measured? What is the control? What does the data show, and what does it not show?
Core concepts
11.1Variables and controls
The independent variable is manipulated, and the dependent variable is measured. Controlled variables are held constant. A negative control should give no effect and shows the method does not create false signals. A positive control should give a known effect and shows the method can detect one.
Ask what the single difference between the experimental and control groups is. If there is more than one difference, the conclusion is confounded.
Key terms: independent / dependent variable · positive control · negative control · confounding variable · replication
11.2Reading graphs and tables
Read the axes, units, and scale (especially log scales) before looking at the data. Distinguish rate (slope) from amount (height). Check whether error bars overlap and how many replicates there were.
Describe the trend, then give the numbers, then give the interpretation. Do not extrapolate beyond the tested range.
Key terms: slope · log scale · error bars · extrapolation · outlier
11.3Basic statistics
Mean, median, standard deviation, and standard error. Chi-square compares observed counts to expected counts: χ² = Σ (O − E)² / E, with degrees of freedom = number of categories − 1 for a simple goodness-of-fit test.
A p-value below 0.05 means results like these would be unlikely if the null hypothesis were true. It does not prove the alternative hypothesis, and it does not measure how large the effect is.
Key terms: null hypothesis · chi-square · degrees of freedom · standard error · p-value
11.4Drawing valid conclusions
Correlation does not establish causation without a controlled manipulation. Necessary and sufficient are different claims. A knockout that abolishes an effect shows the gene is necessary. Showing it is sufficient takes a gain-of-function experiment.
In multiple-choice questions, the correct conclusion is usually the most modest one that the data fully support.
Key terms: correlation vs. causation · necessary vs. sufficient · knockout · rescue experiment · overgeneralization
Practice
Review the concepts above, then complete the practice set. Missed a question? Read the explanation and try a similar problem.
Common mistakes
| Mistake | Instead |
|---|---|
| Choosing the most interesting conclusion. | Choose the conclusion the data directly support. Watch for words like "all," "always," and "proves." |
| Ignoring the control group. | Compare treated samples to the control, not to zero. |
| Reading a log axis as linear. | Each gridline on a log scale can mean a tenfold change. |
Recommended resources
- HHMI BioInteractive: Data Point and data explorer activities Data activities
- Khan Academy: Statistics and probability Videos & practice
- OpenStax Introductory Statistics Textbook (free)
Further reading
- Campbell Biology, the "Scientific Skills Exercise" features throughout the book.
- Past USABO exams (via the official Student Resource Center) for the style of data questions.