Evaluation: (b) the identification of anomalies in experimental measurements

The idea

An anomaly is a reading that does not fit the pattern of the other measurements. To identify one, compare each repeat with the others and with the trend shown on a graph. A value that lies far outside the spread of the rest, or that breaks a smooth line, is suspect. Judge against the uncertainty of the measurement: a reading within the expected scatter is not anomalous. For example, in timing a pendulum, five repeats might give 1.42 s, 1.43 s, 1.41 s, 1.44 s and 1.62 s; the last value is an anomaly because it is far outside the spread. Once identified, repeat that measurement, and exclude the anomaly from a mean only with justification.

What earns the marks

• Defines an anomaly as a reading that does not fit the pattern of the other measurements.
• Compares each reading with repeats and with the overall trend or graph.
• Uses the spread or uncertainty to judge whether a value is genuinely anomalous.
• Identifies a value that lies far outside the expected scatter as suspect.
• Explains that an anomalous reading should be repeated and excluded from a mean only with justification.

Common mistakes and how to put them right

• Treating any value that differs slightly from the mean as anomalous.

  Put it right: Compare the difference with the uncertainty and the spread of repeats.

• Removing an outlier without comment.

  Put it right: Justify the exclusion by showing it lies well outside the expected scatter and state that the measurement was repeated.

• Assuming an anomaly is always a random error.

  Put it right: Consider whether a systematic cause, such as a mis-set instrument, produced the odd value.

Exam tips

• Quote the suspect value and the neighbouring values to show why it is anomalous.
• State the criterion you used, such as lying outside the spread of the other repeats.
• Say what you would do next: repeat the measurement and decide whether to exclude the value.