Outliers analysis identifies specific replicates that are statistically anomalous compared to the other replicates of the sample. Because each individual replicate is compared to the other replicates, there must be enough replicates to establish the “typical” mean and variability of the sample. The minimum feasible number of replicates is 5, but additional replicates give greater reliability and confidence in the statistics.
Many statistical techniques are available to identify outliers. The techniques used by this program are:
- Tukey’s Interquartile Range Fences
- This technique based on the median and quartile values so it is less sensitive to the possible presence of outliers (the median and interquartile range are only minimally affected by an outlier)
- Modified Thompson Tau Test
- This is based on the mean and standard deviation which are sensitive to the presence of extreme points so this technique is more strongly affected by the presence of outliers
Our Outliers program gives clear and easy-to-understand outlier identifications without the requirement of sophisticated statistical interpretation by the user.
Tukey’s method has the added benefit of being able to identify outliers in samples that are subjected to a suite of analyses. For instance, all the data on a replicate can be identified as an outlier based on the result of any analysis on that replicate.
This Outliers program implements enhancements to the normal analysis:
- Set minimum tolerances below which deviations are too small to be considered relevant
- Identify both Outliers and Extreme values
- Simultaneously analyze multiple characteristics of a sample
- Analyze correlated characteristics
- Show the impact on the sample’s summary statistics resulting from rejected outlying replicates
Further information on identifying and dealing with outlying replicates here.
Precision Scientific Software 