Outliers

Dealing with outliers:

1) Identify them and mark them in the RAW data table with a star/asterisk (*). State them in the table caption.

2) Justify, explain and discuss the source/cause of the outliers (use qualitative data).

3) Exclude all the outliers from the data processing (only if they have been justified!).

4) Refer back to them in the evaluation and include suggestions to prevent outliers in the method.

NOTE: If the number of experimental data is very limited outliers should NOT be excluded from the processing. Standard Deviation should NOT be calculated with less that 5 data points.

Outlier(s)

Outliers are abnormal measurements in the raw data. They can be identified by looking at the different results amongst trials. A large standard deviation suggest the presence of outliers. To assess the largeness of the standard deviation, the coefficient of variation can be calculated (see here).

Outliers MUST be identified, marked in the raw data table and explained/justified: the cause/reason of the outlier should be discussed. Sources of outliers may be variability in the measurements or the organisms, but they may also may indicate experimental errors (random error?, species variability?, mistake in measurement). After a discussion about the possible source of the outlier(s) is presented, data processing should be done excluding outliers. If the outlier is a random error, why would you include it in your processed data?

To make the analysis more complete, two data processing and two graphs could be shown, with and without outliers. In order to prevent the report being lengthy, the data processing and graph of the data that includes the outlier should be shown in a separate appendix at the end of the lab report (titled data analysis including outliers) and referenced in the text at the beginning of the analysis section.

Data range and outliers

The range is the measure of the spread of data: the difference between the largest and the smallest observed values. If one data point was unusually large or unusually small, this very large or small data point would have a big effect on the range. Such very large or very small data points are called outliers. If an experiment is measuring the change in height of plats (as an independent variable is modified) and one of the plants dies early and had an unusual low height, that data point would be considered to be an outlier. In a lab report, it is acceptable to exclude an outlier from data processing, but it is important to declare it and explain why it was excluded. The best option would be to show the data processing with and without the outliers, explaining how the outliers were identified and why were them removed. Outliers may make a big difference in the graphic presentation of data, so a graph without outliers is strongly encouraged.