How does statistiXL compare to Excel’s Analysis ToolPak?
26 September 2014
statistiXL complements and expands on the statistical tools provided by the Analysis ToolPak that comes with Excel. statistiXL provides a greater range of more sophisticated analyses, including many that are not included in the ToolPak, and provides greater functionality where similar tests are included in both. Below I run through the modules provided by statistiXL and list the differences between it and the ToolPak.
Starting with the modules that overlap between statistiXL and the ToolPak:
- Analysis of Variance – Both statistiXL and the ToolPak support Analysis of Variance but the ToolPak only supports 1 and 2 factor ANOVA whereas statistiXL is only constrained by the size of your Excel Spreadsheet or your PCs memory. The ToolPak also only supports full factorial ANOVA whereas statistiXL lets you specify user defined models enabling you to analyse experimental designs, such as Split Plot , Latin Square or Nesting. statistiXL also allows you to control for one or more covariates into your analysis (i.e. Analysis of Covariance) and can perform post hoc test on your results to show where differences between groups lie.
- Correlation – The overlap between the correlation modules provided by statistiXL and the ToolPak is very limited. Excel’s ToolPak simply tabulates the correlation coefficient for adjacent columns of variables. statistiXL goes far beyond this, providing Simple, Multiple, Partial and Canonical Correlation. Coefficients are provided along with appropriate tests for significance. statistiXL can also produce appropriate graphs for each analysis where applicable (e.g. scatter plots for Simple Correlations of plots of the variates for Canonical Correlations).
- Descriptive Statistics – The ToolPak provides a basic Descriptive Statistics module for basic linear descriptive statistics. statistiXL provides an enhanced module with additional statistics (such as Percentiles and the Coefficient of Variation) and the ability to output Box and Whisker Plots and Error Bar Plots. statistiXL also provides separate modules for descriptive statistics for circular variables (such as time, compass point measurements etc) and for frequency analysis.
- Linear Regression – Both the ToolPak and statistiXL can perform Simple Linear Regression on a pair of dependent variables, outputting an F test of the relationship and various plots, such as scatter plots and Normal Probability Plots. statistiXL however extends Linear Regression to cover Weighted Regression, Multiple Regression and also Forward and Backward Stepwise Regression.
- t Tests – The ToolPak supports 2 Sample Paired and Unpaired tests and tests can be performed assuming equal variances or unequal variances but the user must specify which to use. statistiXL also supports Paired and Unpaired 2 Sample t Tests. The user can specify whether variances are assumed to be equal or not or the appropriate method can be chosen automatically based on the results of an F test. statistiXL also supports single sample t Tests (i.e. tests against a hypothesised mean) and multivariate t Tests.
In addition to these modules, statistiXL provides a range of other tools not included in the Analysis ToolPak. The following list is taken from the Features section of our website:
- Clustering – Hierarchical clustering of binomial, quantitative and mixed datasets is supported as is clustering based on a predetermined distance matrix. A wide variety of similarity/distance estimates and clustering methods are available and the resultant clustering strategy can be graphically displayed as both a text based and/or graphical dendrogram.
- Contingency Tables – Both 2-way and multi-way contingency data can be analysed.
- Discriminant Analysis – Both Grouping and Classification methods of Discriminant Analysis are supported. For Grouping Discriminant Analysis, scatterplots of case scores can be produced for each pair of components. For Classification Discriminant Analysis, an alternate dataset can be classified based on the discriminant functions determined for the primary set.
- Factor Analysis – Factor Analysis can be performed on either the correlation or covariance matrix of the raw data set. A variety of component extraction and rotation methods are supported and both scree and scatterplots of case scores can be produced.
- Goodness of Fit – A wide variety of tests for the Goodness of Fit of datasets to theoretical distributions are provided including those for Binomial, Circular, Normal, Poisson and Uniform distributions. The level of fit to user specified distributions can also be calculated.
- Nonparametric Tests – Numerous Nonparametric Tests are supported including Friedman, Kruskal-Wallis, Mann-Whitney, Mood’s Median, Sign, Spearman, Wald-Wolfowitz and Wilcoxon Paired-Sample tests.
- Principal Components – Principal Component Analysis is provided as a means for the reduction of large multivariate data sets into simpler structures. Scree plots and Scatterplots of case scores can be produced.
The above provides a simple overview of some of the major differences between statistiXL and the Analysis ToolPak. statistiXL provides a range of tools and options that 1) are not available natively within Excel and 2) will address a wide range of analytical questions often posed by scientific research.