- LAB Fit
- GraphPad Prism
- SimFit (free)
- SciDAVis (free; really more like GraphExpert)
- TableCurve 2D
- MagicPlot ($198)
CurveExpert Professional ($79) is the clear winner in terms of pricing.
Only LAB Fit ($85) can compete, in terms of price, with either CurveExpert Pro.
DataFit is quite expensive ($299), especially since CurveExpert Professional exceeds its capabilities.
GraphPad Prism ($200 subscription per year) is also well out of CurveExpert Pro's price range.
CurveExpert Professional supports Windows, Mac, and Linux, all from the same license.
Out of all of the packages, CurveExpert Professional is the only one that supports Windows, Mac, and Linux. This gives you maximum flexibility, as it is common for multiple platforms to be used in conjunction with any given task.
Of the competition, only GraphPad Prism ($595) supports any other platform other than Windows.
CurveExpert Professional does not limit the number of independent variables.
All of the products support multiple independent variables. CurveExpert Pro does not limit the number of independent variables in any way. In all but the most exceptional situations, six independent variables should certainly be sufficient.
The graphing quality of CurveExpert Professional is unmatched by any of the competitors. Graphs are fully customizable, semi-interactive, and fully anti-aliased for publication quality images.
All packages provide some sort of capability for creating your own models for nonlinear regression. All are reasonably easy to use, and are based on a text representation of the equation that you would like to model.
CurveExpert Professional differentiates itself by automatically rendering your equations into a graphical format, and also provides an extremely powerful interface for writing custom models in Python. With these advanced user models, arbitrarily complex models can be expressed.
Clearly, DataFit has the greatest number of built-in models. The reason for this, however, is that DataFit does not provide automatic initial guesses for any of its models. This causes models to fail to calculate that normally would in CurveExpert Pro, or forces you as the user to always come up with initial guesses for the built in models.
The philosophy of CurveExpert Pro is to provide families of models that cover the common data shapes (S-shaped, steadily rising/falling, etc.). High quality automatic initial guesses are always provided for the built-in models, to maximize their effectiveness. Past these shapes, you are free to customize a model that fits your needs. It is not useful to the user to provide endless permutations of the same model in order to increase the model count.
The usability metric is highly subjective. This measure takes into account the ease with which data may be imported from external sources into the software, created from scratch by the user, and how difficult it is to create and use results from the software.
- CurveExpert Pro can read files from practically any source; the file reading is extremely flexible and customizable, and can automatically detect where the data is in the file. Data can be entered manually (if desired) in a power Excel-like spreadsheet with cut/paste capability, or data files can be dragged and dropped straight into the application. Full localization is supported, so European customers do not have to fight with commas while reading, copying, or pasting data.
- LAB Fit requires being an admin to install, and importing files is fairly difficult. The interface is extremely hard to use, and no spreadsheet is present for data entry (data must be entered into a forms-based interface).
- DataFit can certainly read data files from other sources, but requires much more interaction from the user, in order to skip non-data items properly. Columns in the data file must be manually assigned to variables, which can be very time consuming. The output statistics for a curve fit are thorough and clear.
- Graphpad Prism is more of a heavyweight application which supports many more types of statistical analyses than any of the other packages tested. It is very much like Mathcad in its interface approach. With the larger featureset, though, comes a much higher price tag, and a steeper learning curve to be productive.