GraphPad Curve Fitting Guide



Previous topic Next topic No expanding text in this topic  


Previous topic Next topic JavaScript is required for expanding text JavaScript is required for the print function Mail us feedback on this topic!  

The outliers table

Choose to exclude or count outliers as part of nonlinear regression on the Method tab.

If you do this, one of the results tables will list all outliers.

Keep in mind an important distinction:

If you chose to count outliers , then the outliers are still included in the fit.

If you chose to exclude outliers , then the outliers are ignored by the fit, but are still included on the graph.

In both cases, the aggressiveness of the outlier hunt is determined by the ROUT coefficient Q that you enter.

What does it mean to be an outlier

Prism uses the ROUT algorithm, using the value of Q that you can adjust to decide when to declare a point to be an outlier. Some considerations:

The first thing to do is make sure it isn't simply a problem in data entry. If so, fix it.

Review the list of situations where the outlier identification procedure should not be used.

Be alert to biological variation. In some experimental situations, outliers can only arise from lab mistakes. These should be excluded. In other situations, an outlier can be the result of biological variation. You may have discovered a new polymorphism in some gene. It could be missing a big opportunity if you exclude and ignore those values.

What to do once you have identified an outlier

After you have identified one or more outliers, you have several choices.

While outliers are ignored by the nonlinear regression calculation, they are still plotted on the graph. If you want the outlier to be entirely removed from the graph, including error bars on the graph, go back to the data table and exclude the outlier(s). Prism will then refit the curve, and replot the graph without any outliers.

If you want the outlier to remain on the graph, simply polish the graph Prism provides. Keep in mind that the outliers are a dataset, so you can separately adjust the size, color and symbol of outliers and all the data points. Use the "Datasets on graph" tab of Format Graph to adjust the back-to-front order of data sets to keep the outlier in front of the other data sets. The dialog shows the datasets in back to front order, so you want to keep the outlier data sets lower in the list.

If you want to remove the special formatting of outliers on the graph, use the "Datasets on graph" tab of Format Graph. Select the outlier dataset(s) and click Remove. The outlier will still be plotted as part of the full data set, but will no longer be plotted separately as an outlier data set.

If you asked Prism to count the outliers (Diagnostics tab) but now wish to fit the data with the outlier ignored, go back to the nonlinear regression dialog and choose "Automatic outlier elimination" (on the Fit tab).