Overview
If you are accustomed to Prism 8, you’ll find it easy to switch to Prism 9. Nothing fundamental has changed. You’ll just see more options in many places. If you have large data tables, you’ll also notice that Prism runs more quickly.
Highlights
•New analyses: Principal Component Analysis (PCA) and Principal Component Regression (PCR)
•More automation: Automatically add multiple comparison results to graphs ("Stars on Graph")
•New Graphs: Bubble plots and other multiple variables graphs from multiple variables data tables
•New Graphs: Estimation plots automatically generated from t tests
•New Graphs: Actual vs Predicted plot from nonlinear regression
•Expanded analyses: New options for Multiple t test analyses (paired, nonparametric, and more)
•Expanded analyses: Interpolation from Multiple Linear Regression
•More data: Increased data table limits
•More data: Text variables and variable types (continuous, categorical, label) for multiple variables data tables
Additional Details
•A data table can now have 1024 data sets [letters A...AMJ] and 512 sub-columns (up from 512 and 256).
•A graph can now plot 1024 data sets.
•Speed. Many analyses are faster
•No longer runs on 32-bit Windows, as 64-bit Windows has been standard for a decade.
•You can now enter text variables into multiple variable tables.
•Each variable (column) is now identified as a continuous (numeric), categorical, or label variable
•Multiple and logistic regression can now fit models where one or more of the predictor (independent) variables is entered as text and designated as a categorical variable.
•Prism knows how to fit models where a categorical variable has three or more levels (possible values). It creates new variables (one less than the number of levels) to include in the model, using the reference level method.
•As you enter (or import) data, Prism defines each variable (column) as continuous, categorical or labels. You can override this automatic decision as needed.
Principal component analysis (PCA)
•PCA is a method used to analyze data with many variables. It determines a number of “components” (each a linear combination of the original variables) that explain much of the variation. For example, it could reduce a data set with 15 variables down to a data set with four components that explain most of the variability. PCA is mostly used as a tool in exploratory data analysis and for making predictive models. It is often used to visualize genetic distance and relatedness between populations.
•You can now create five residual graphs (including the new Actual vs Predicted graph). Before you could only plot one. Read more
•User-defined equations are evaluated much more quickly (and more accurately)
•When you fit a model defined by a differential equation, one of the parameters fit is Y[X0], the value of Y when X equals X0 . In Prism 8, you could set the value of X0 when defining a user-defined differential equation. With Prism 9, X0 is a parameter that must be set to a constant value, but you can set it to any value in the nonlinear regression parameters dialog.
The multiple t test analysis in Prism 8 only could perform the unpaired (two-sample) t test, one per row. Now, it can also perform (one per row), these analyses:
•Unpaired t-tests with Welch correction
•Paired t-tests
•Ratio-paired t-tests
•Nonparametric unpaired Mann-Whitney tests
•Nonparametric unpaired Kolmogorov-Smirnov tests
•Nonparametric paired Wilcoxon tests
Multiple linear and logistic regression
•Both multiple linear and multiple logistic regression can fit a model where one (or more) predictor (independent) variable is categorical with three or more levels. Prism automatically creates new variables ("dummy variables", one fewer than the number of levels) to use in regression. Options available to specify the "reference level" for each categorical variable in the model. Read more (multiple linear regression) or Read more (multiple logistic regression)
•Interpolate outcome (dependent) variable values from data in the data table or by using specified values for each predictor in a model with multiple linear regression.
•The first tab of the dialog, where you define the model, has been redesigned.
•Multiple regression and logistic regression can output a covariance matrix, and create a heat map, showing how the explanatory variables correlate with each other.
•Calculate the confidence interval of mean with any confidence interval level (Prism 8 only computed 95% CI).
•Calculate medians with "no errors", "quartiles", "min / max", or "percentiles"
•Calculate geometric means with "no errors", "geometric SD", “CI”
•T tests. Plot the results of unpaired and paired t tests with an estimation plot. This plots both raw data and the effect size (difference between means) with a confidence interval. Read more
•Two-way ANOVA. Choose whether you wish to include an interaction term in the model. Previous Prism versions always fit the interaction term. Now it can fit a main-effects only model.
•Two-way ANOVA can now be calculated even if there is no data for some row/column combinations (but only if you choose not to fit the interaction term).
•You can now ask Prism to ignore any row where any value is missing or excluded. With this option selected, all the correlation coefficients in the matrix are computed from the same set of rows. Read more
•Visualizing pairwise comparisons. Automatically add significance stars to graphs with a single click after performing an appropriate analysis on the data. Read more
•More color schemes (semi-transparent variants of popular color schemes).
•New Insert ..Character>Unicode Symbols...
•Improved Prism's "Insert Character Dialog" to use Unicode rather than Symbols font for Greeks/Maths/European characters. This makes Prism graphs more compatible with other platforms and apps.
•Graphs from multiple variable tables. Variables ecode position (X- and Y-coordinates), and additional variables (numerical or categorical) can be selected to encode color and size. Read more
Features added since 8.0 you may have missed
Logistic regression
•Model binary probability (e.g. yes/no, pass/fail) with a single or multiple explanatory variables. There are two new analyses, simple logistic regression to analyze data in an XY data table with one explanatory (independent) variable, and multiple logistic regression to analyze data in an Multiple Variable data table with any number of explanatory (independent) variables (added in 8.3).
•Prism reports confidence interval for parameter "X at 50" for and provides the option to plot confidence bands when performing simple logistic regression (added in 8.4)
Nonlinear regression
•All concentration-response curves are now available in two forms, with X as concentration or X as log(concentration), (added in 8.2).
•Improved Prism's ability to identify poor fits. Instead of identifying an entire fit as "ambiguous", Prism provided the option to identify specific parameters as "unstable" (introduced in 8.2, made default in 9.0).
ANOVA
•Nonparametric one-way ANOVA (Kruskal Wallis test) followup. Dunnett’s T3 test compares the mean rank of each column with the mean rank of every other column (added in 8.1)
Results sheets
•The appearance of results tabs improved (added in 8.2)
•Prism re-introduced the ability to group Data and Results tables together in the Navigator... again! (This feature was removed in 8.0, but re-introduced in 8.2)
Graphing
•Automatic creation of heat map of correlation coefficients after performing Correlation Matrix analysis (added in 8.1)
•Prism now offers three ways to arrange dots in scatter graphs. The default choice avoids the “smiles” arrangement (added in 8.2).
•New color schemes (Colorblind Safe, Prism Light, Prism Dark, Floral, Waves, Pearl, Starry, Viridis, Magma, Inferno, Plasma) (added in 8.1 and 8.4)
•Violin plots in Prism 8.0 were capped to always extend from the smallest value to the largest value. Now Prism offers a choice to extend the violin plots beyond the range of the data to predict the range of the population (added in 8.4)
