lesson
Updated 6 days ago · 1 view
If you measure the time for a magnesium ribbon to dissolve in acid three times and record 24 s, 25 s, and 82 s, your stopwatch didn't discover a new law of physics on the third run.
In scientific experiments, we collect repeated measurements, called replicates, to spot random errors and find the most reliable central value.
What makes a rogue data point an error rather than natural variation, and what should you do with it?
Identifying and Discarding Anomalies
An anomaly (or outlier) is a recorded value that lies far outside the expected range or pattern of concordant replicates.
📊A clean visual card comparing 4 replicate measurements of a reaction time: Trial 1 (24.2 s), Trial 2 (24.6 s), Trial 3 (82.0 s with a bold red cross-out badge saying 'ANOMALY - DISCARD'), and Trial 4 (23.8 s). Show an arrow pointing from the 3 valid trials to a 'Valid Dataset: n = 3' container. Use card background #ffffff, border #e6e6e6, text #1e2945, accent red #ef4444 for anomaly, and accent green #10b981 for valid trials.
Whenever you spot an anomaly caused by experimental error, you must discard it completely before running any statistical calculations.
Now that the anomaly is removed, how do we combine the remaining numbers into a single reliable average?
Calculating the Arithmetic Mean
The arithmetic mean represents the central balance point of a dataset, calculated by adding all valid numbers and dividing by the count of those valid trials (n).
📊An interactive-style worked example card for calculating the mean. Formula shown at top: Mean = (Sum of valid replicates) / n. Step 1: Add valid replicates: 24.2 + 24.6 + 23.8 = 72.6. Step 2: Count valid replicates: n = 3 (highlighted with a badge: 'NOT 4!'). Step 3: Compute 72.6 / 3 = 24.2 s. Highlight final answer in a blue callout box. System font, #1e2945 text, #f0fdf4 green highlight for sum, #eff6ff blue highlight for final mean.
A common mistake on science exams is dividing by the original total number of trials rather than the number of remaining valid trials.
If you discard 1 anomaly out of 4 trials, your new divisor is n=3, never 4.