Accuracy and Precision are properties of an experiment which are often conflated – not least because they are often synonymous in everyday conversation. In science, however, they have separate, well-defined meanings. Of course, this also applies to bioassay, so we thought we’d take a look at the difference here.
Accuracy is the ability of an experiment to return an answer which, on average, is close to the “true” value.
Imagine a temperature-controlled laboratory in which the temperature was set to 21.5 degrees Celsius. A thermometer would be considered accurate if it, over a series of measurement, returned an average temperature of 21.5 degrees.
If, instead, it returned an average temperature of 22.5 degrees, we might consider the thermometer inaccurate. Many experiments show an inaccuracy in a single direction – here, the inaccurate average is greater than the true temperature – so we often say that an inaccurate experiment is biased.
Key Takeaways
-
Accuracy is the degree to which experimental results average close to the true value. For instance, a thermometer consistently measuring the actual temperature accurately over multiple trials is considered accurate.
-
Precision is the consistency or repeatability of measurements, indicating how much the results vary. A thermometer that measures temperature to two decimal places is more precise than returning only integer results.
-
Bias is a systematic deviation from the true value in a particular direction. If an experiment consistently overestimates or underestimates the true value, it is said to be biased.
Precision, on the other hand, is a measure of the variability of an experiment – how spread out are the possible results of the experiment? For our laboratory thermometer, a device which measures to, say, 2dp (i.e. 21.51 degrees, 21.75 degrees) would be considered more precise than a device which only measured to integer values (i.e. 21 degrees, 22 degrees).
In many cases, there is little correlation between accuracy and precision. In the images below, we have represented the “true” assay result with a solid white circle. The measured results are shown by faded white circles. These images show the four broad cases of assay accuracy and precision.
Low Accuracy, Low Precision
Not good: The average estimate result is nowhere near the true result i.e. it is biased and imprecise.
More tests? No, more tests just gives a better estimate of the biased result
Plan: You need to investigate and correct the cause of the bias, and the variability
Low Accuracy, High Precision
Good in parts: The average result is precise, but not a good estimate of the true value. Precise, but biased.
More tests? No! This would just provide an even more precise estimate of the wrong result.
Plan: Investigate bias
High Accuracy, Low Precision
Good in parts: The average result is a good estimate of the true value, but individual results are variable. Accurate, but not precise.
More tests? Yes! More data will improve the precision.
Plan: Do more tests and / or decrease variability.
High Accuracy, High Precision
Very good: All is well. The estimate you will get from the average of these tests is both accurate and precise.
Plan: Nothing to improve!
Comments are closed.