JMP (pronounced “jump”) is an interactive statistical visualization tool. JMP is designed to provide data analysis with unique statistical visualization. Each kind of analysis begins with graphs and plots, and each plot is accompanied with the appropriate statistical summary tables. So use this demo version to jump right in and see what it can do.
The demo version of JMP 3.0 is fully functional except for read and write capability. It only opens the data tables in the SAMPLE DATA folder and it does not save data tables. JMP Demo will not import data from the clipboard, or from the Import command. JMP Demo is free, and was developed for you to evaluate JMP and to learn how to use it. It may be freely distributed, given to friends, and placed on bulletin boards, but it may not be distributed as part of commercial software. Please do not distribute the application without its help file, sample data tables, or this documentation file. We want people to be able to understand what they are getting.
This document gives you ‘getting started’ directions. It begins with a quick example and follows with a description of the main features of JMP. Don’t be afraid to experiment with the analysis and graph platforms. Wherever you see an icon, click to see what it has to offer. Try modifier keys like shift, command, and option to see what effect they have. Use the “Table Info” command on the sample files for suggested analyses.
The demo also includes samples from JMP’s help system.
• Use the ? tool in the Tools menu for context-sensitive help.
• Choose the About JMP item on the Apple menu and click whenever you see (?).
• In particular, select the statistical guide (the Guide button from the About JMP item on the Apple menu). This activates a scrolling list of available statistical analyses. When you click an analysis, directions for doing the analysis appear.
• Select HELP from the star pop-up menu on the lower left border of each analysis window.
A sample of JMP balloon help is available if you are running Macintosh System Software 7.0.
The following example is designed to take you on a short guided tour through a JMP session. Follow the steps to see a three-dimensional spinning plot.
Begin a JMP session by opening the JMP file called COWBOY HAT. This spreadsheet consists of three numeric columns. x and y are rectangular grid coordinates, and z is a function of x and y.
To plot the three columns of information from the COWBOY HAT data table, choose Spin from the Analyze menu. A dialog prompts you to select columns for the Spin platform. Select the x, y, and z columns from the list of available columns on the left side of the dialog and click the Add button. When you click Done, the spin platform appears and the points in the COWBOY HAT data table are plotted on the spin platform.
Next, select the hand tool from the Tools menu. Position the hand on the cowboy hat spin plot, hold down the mouse button, and move the hand about. The cowboy hat moves in three dimensions. The plot can also spin by itself. Hold down the shift key and give the plot a push with the hand. To stop the spinning plot, click in any JMP window.
The spin control panel showing on the left of the plot contains twelve small icons. These icons control the spin plot. The six directional arrows are buttons that control the spin of the three dimensional image. If you click one of these arrows, the plot spins in the direction indicated. As long as you hold down the mouse button, the spin persists. However, you can shift-click a directional arrow to make the plot spin continuously.
Click the triangle icon to see a pop-up menu of Display Options. These options let you tailor the spinning plot. Complete your experiment with the spinning cowboy hat by trying out a few of the display options.
To exit JMP, select Quit from the File menu.
2. THE JMP APPROACH TO STATISTICS
This section introduces you to the JMP analysis platforms and presents the basic navigational instructions on how to use them.
Each command from the Analyze and Graph menus launches a platform. A platform is an interactive window that you can use to analyze data, work with points on plots, and save results. The platform persists until you close it, or close the data table to which it is linked.
Before you launch a platform you give JMP information about the data you want to analyze:
• Data type
Each column has a data type of numeric or character. You can specify the data type in the Column Info dialog of each column
Numeric…numeric columns can contain only numbers and decimals.
Character… character columns can contain any kind of information. All values, including numbers, are treated as character values.
• Modeling Type
The modeling type assigned to a variable can be continuous, nominal, or ordinal. To assign a level you can use the pop-up menus at the top of each column on the data table window or respond to the dialog given by the Column Info command in the Cols menu.
The modeling types are not just descriptive tags. They tell each analysis platform how to analyze and graph the data. They are prescriptions for treating the values in a model. Furthermore, there is a difference in modeling type treatment according to whether the variable is a response (Y) variable or a factor (X) variable. Do not hesitate to change the modeling type in order to study the same variable in different ways.
The modeling types used by JMP are defined the following way:
--Continuous… For continuous columns, the numeric value is used directly. Often the variable is measured on a continuous scale.
--Ordinal… An ordinal value is not used directly, but only as a name. The values can be either numeric or character. However, the values are interpreted by JMP as having an order.
--Nominal… A nominal value is just a name, and the values are interpreted as having no particular order in the analysis conducted. The values can be either character or numeric. Numeric values are treated as discrete values.
• Variable Roles
You choose X or Y roles for the variables depending on how you want JMP to treat them. A column assigned the role of Y is a response or dependent variable. An X column is a factor variable that fits, predicts, or groups a response. You can mark columns as X or Y in the data table before you launch the platform using the pop-up menu at the top of the column.; or, if no variables are assigned X and Y roles on the spreadsheet, each analysis platform prompts you with a role selection dialog.
3. Navigating the Analyze and Graph Menus
——Analyze menu: Analysis Platforms
You launch an analysis platform from the Analyze menu according to the way you want X and Y variables treated together. The following list briefly describes each of the eight Analyze commands. Also, the HELP button for each Analyze command is active from About JMP in the Apple menu, and on the analysis display platforms.
• 1. Distribution of Y’s describes the distribution of each Y column using histograms and other graphical and textual reports. If a column has continuous values, the graphical display consists of a histogram, outlier box plot, and quantile box plot. If a column has nominal or ordinal values, the graphical display consists of a histogram and a divided (mosaic) bar chart showing a bar for each level of the ordinal or nominal variable. The Distribution of Y platform also has normality tests, an option to test a mean against a specific value, and can perform a capability analysis.
• 2. Fit Y by X describes each pair of X and Y columns. The displays and reports vary depending on the levels of measurement you specify for the X and Y columns.
a) When both X and Y are continuous variables, the Fit Y by X performs a regression analysis with a variety of fitting options such as linear, polynomial, smoothing spline, and others. You can request as many fits as you want, which gives you an interactive exploratory fitting platform.
b) When X is nominal or ordinal and Y is continuous, the Fit Y by X does a one–way analysis of variance (ANOVA) platform with a variety of multiple (post hoc) means comparisons. The one-way ANOVA is represented graphically by a dot plot with options to see means diamonds, box plots, and means, standard deviations, and error bars for each group.
c) When Y is nominal or ordinal and X is continuous, the Fit Y by X platform gives a logistic regression and displays a family of logistic probability curves. If the response has only two levels, you can do inverse prediction.
d) When both Y and X are nominal or ordinal variables, The Fit Y by X platform gives a two–way contingency table analysis with a mosaic plot of the table, an analysis of log likelihood table with Pearson and Likelihood ratio chi–square statistics, tables of frequency counts and percents, and an optional correspondence analysis. If both X and Y have only two levels, a Fisher's exact test is performed.
• 3. Specify Model displays a dialog that defines models automatically, or lets you build specific model effects and specify error terms. It is used for multiple regression analysis, stepwise regression, logistic modeling, ,analysis of variance, analysis of covariance, analysis of categorical linear models, multivariate analysis of variance, log-linear modeling, analysis of proportional hazard models, and to search for D-optional designs. Results are presented with tables, leverage plots, and profile plots. In addition you can request custom tests, contrasts, power details and Type II Sums of Squares , do inverse prediction, and request a Durbin-Watson test. You can save predicted values with the prediction formula, residuals, predicted and individual confidence values, hat values, leverage pairs, and Cook’s D.
• 4. Nonlinear Fit fits nonlinear models, which are models that are nonlinear in their parameters. The Nonlinear Fit command launches an interactive fitting facility. You orchestrate the fitting process as a coordination of three important parts of JMP: the data table, the calculator, and the nonlinear fitting platform.
• 5. Correlation of Y’s describes relationships among multiple Y’s with Pearson correlations, inverse and partial correlations, nonparametric correlations, a scatter plot matrix, and multivariate outlier plots, and can compute principal components and show factor analysis information.
• 6. Cluster hierarchically clusters rows in a JMP table. You can choose from five different clustering methods, see the clustering history, a dendogram (tree diagram) of the clusters, automatically color and mark clusters in plots, and save the distances between clusters in a data table.
• 7. Survival computes life-table estimates of survival functions using the Kaplan-Meier survival method for one or more groups of right-censored data.
Note ====> You can use the Proportional Hazard fitting personality found in the Fit Model dialog for fitting failure time response with right-censoring and covariates, and the nonlinear platform for failure time models with a specific survival distribution, using a loss function to produce maximum likelihood estimates.
Important===> For a description of how to do some specific types of analyses, choose About JMP from the Apple menu. Then click the Guide button. You will see a scrolling alphabetical list of statistical methods. Click the method you want to do to see a description of how to specify the variable roles, their modeling types, and which platform to use.
Graph Menu: Graph Platforms
• 1. The Bar/Pie Charts platform displays a bar, line, or pie chart for numeric each variable (Y column) you specify, showing the value for each level of the X column.
• 2. The Overlay Plot platform displays an overlaid bar or line plot of a single X column and all the numeric Y columns you specify. You also have the option of separating the plots.
• 3. Spin produces a three-dimensional view of data and an approximation to higher dimensions through principal components. The plot is a spinable display of the values of any three numeric columns you choose. You can see a factor–analysis–style rotation (varimax rotation) of the principal components to form orthogonal combinations that correspond to directions of variable clusters in space.
• 4. The Pareto Chart platform gives a bar chart or series of related charts for a process variable (Y column) and up to two classification variables (X columns).
• 5. The Control Charts platform displays control charts, sometimes referred to as Shewhart charts, which are a graphical and analytical tool to help you decide if a process is in a state of statistical quality control. Types of control charts are charts for variables (mean, standard deviatio, range), individual measurement charts, charts for attributes (P, NP, C, U), moving average charts (EWMA, UWMA), and cusum charts
• 6. The Contour Plot platform constructs contours of a response in a rectangular or a triangular coordinate system. You specify two numeric X variables for the X and Y axes, and one numeric Y variable as the response.
• 7. The Ternary Plot is a way of displaying the distribution and variability of three-part compositional data such as the proportion of sand, silt, and clay in soil, or the mixture proportions of three chemical agents in a trial drug.
An important feature of Bar/Pie Charts, Overlay Plots, and Control charts is that they update dynamically as values are added to the data table.
4. THE CALCULATOR—A WAY TO CREATE NEW COLUMNS
The JMP calculator is a powerful tool for building formulas that calculate column values. JMP formulas can use information from existing columns in the data table, built-in JMP functions, and constants. Formulas can be simple assignments of numeric, character, or row state constants or complex evaluations based on conditional clauses.
When you create a formula for a column, that formula becomes an integral part of the data table. The formula is stored as part of a column’s information when you save the data table, and it is retrieved when you reopen the data table. You can examine or change a column’s formula at any time by opening the calculator window.
The following calculator example gives you a quick look at the basic features of the calculator. For this example, open the JMP file called BIG CLASS included with JMP Demo. It has a column called weight. To begin, create a new column by selecting New Column from the Cols menu. The dialog that appears when you select New Column lets you set the new column’s characteristics. Type a new name, Std. Weight, in the Col Name area and select “Formula” from the Data Source pop-up menu. When you click the OK button in the New Column dialog, the calculator window opens.
Next, enter the formula that standardizes the weight values:
• 1. While the initial missing term is highlighted, click the column named weight in the column selector list.
• 2. Click the minus button in the keypad.
• 3. While the new missing term is highlighted, click weight again.
• 4. Click in the function browser topics list and scroll down the left side to locate Statistical functions. Click this topic to see a list of statistical functions on the right half of the function browser. Click “Mean” in this list.
• 5. Click the right parenthesis in the keypad to highlight the entire expression.
Note: Step #5 is important because the currently highlighted expression becomes the argument for the next function you choose. The argument is grouped with that function such that the function cannot be highlighted separately.
• 6. While the entire expression is highlighted, click the division button in the keypad.
• 7. Choose weight again from the column selector list.
• 8. While weight is still highlighted in the denominator, choose Std. Deviation from the right half of the function browser.
Now close the calculator window to see the column fill with values. If you change any of the weight values, the calculated Std. Weight values are recomputed.
5. MANAGING ROWS AND COLUMNS (THE ROWS AND COLS MENUS)
JMP data are organized in memory as rows and columns of a table which is referred to as the data table. The columns have names and the rows are numbered. An open data table is kept in memory, and you communicate with it through an active spreadsheet window. You can open as many data tables in a JMP session as memory allows. JMP data tables can be stored in files on disk, sometimes called documents.
Commands in the Tables, Rows, and Cols menus provide data handling operations ranging from standard Macintosh tasks to sophisticated editing and database manipulation.
——Selecting Rows and Columns
The selection of rows and columns in a JMP spreadsheet is done by highlighting them. To highlight a row, click the space that contains the row number. To highlight a column, click the background area above the column name.
To extend the selection of rows or columns, drag across the array or shift-click the first and last row or column range. Use command-click to make a discontiguous selection. To select a subset of the spreadsheet, drag the mouse to form a rectangular box. You can select rows with the excluded, hidden, or labeled row state characteristic using the Select submenu in the Rows menu.
——Adding and Deleting Rows
To add rows to a JMP spreadsheet, use the Add Rows command from the Rows menu or double-click in the triangular Rows area in the upper left corner of the spreadsheet. If a column is computed using the calculator, then its new cells are automatically filled with values. Otherwise, the cells have missing values. You can also add new blank rows by double-clicking anywhere in the data area beyond the last non blank row.
To delete rows from the spreadsheet, select the rows you want to delete and choose the Delete Rows command from the Rows menu. If you mistakenly delete rows, immediately select the Undo Delete Rows command from the Edit menu.
——Adding and Deleting Columns
To add column to a JMP spreadsheet, use the New Column command from the Cols menu or double-click in the triangular Cols area in the upper left corner of the spreadsheet. The dialog then asks you to name the new column and provide column characteristics.
You can key data into the new column, or choose a data source from a pop–up menu in the New Column dialog. If you select “Formula” from the data source pop–up menu, the calculator window opens. You can use this window to construct a formula that computes values for the new column. The formula can include existing columns, functions, and constants, and it can use conditional logic.
Columns can also be added to the spreadsheet by using save commands within the statistical platforms. These commands are available in the save ($) pop-up menu, which accompanies each analysis window.
To delete columns from the spreadsheet, select the columns you want to delete and choose the Delete Columns command from the Columns menu. If you mistakenly delete columns, immediately select the Undo Delete Columns command from the Edit menu.
——Row States
Row states are special characteristics associated with a row. They are used to distinguish subsets of your data, exclude data from analyses, and customize the appearance of graphical displays. Row state commands are located in the Rows menu and affect only highlighted rows. After row states are assigned, they can be saved permanently with the data table in a special row state column. The six row states are defined in the following list.
• Exclude/Include is a toggle that excludes selected rows from statistical analyses.
• Hide/Unhide is a toggle that suppresses the display of points in all scatter plots.
• Label/Unlabel is a toggle that labels points on all scatter plots.
• Colors lets you assign any colors to highlighted rows (on color monitors).
• Markers assigns a character to replace the default dot in scatterplots and spinning plots.
• Selection occurs when you click a row and the row number area highlights. You can also use the Select commands in the Rows menu.
To save row state assignments, create a new column and select “Row State” in the Data Type pop–up menu. Then Select the “Copy from Row State” command in the » pop–up menu at the top of the new row state column. This copies the active row states showing next to the row numbers to the column. To activate a row state column, select the “Copy to Row State” command in the « pop–up menu at the top of the row state column.
6. DATA TABLE MANAGEMENT (THE TABLES MENU)
The Tables menu commands create new data tables by modifying the active data table or by combining several tables.
• Group/Summary—creates a new table called a summary table that has one row for each level of a grouping variable you specify, and can have columns of summary statistics.
• Subset—creates a new data table that is a subset, which consists of selected rows and columns from the active spreadsheet.
• Sort—sorts a JMP data table by one or more columns.
• Stack—creates a new data table by stacking specified columns from the active data table into a single new column.
• Split—creates a new data table by splitting one or more columns to form multiple columns.
• Transpose—creates a new table that is the transpose of the active data table. The columns of the active table are the rows of the new table, and its rows are the new columns
• Join—creates a new data table by merging (joining) two tables side by side.
• Concatenate—creates a new data table from two or more open data tables by combining the tables end to end.
• Attributes—creates a new data table called an attribute table from the active data table, called the source table. There is a row for each column in the source table. The source table columns contain table attribute information.
• Design Experiment—creates table of experimental runs for the design you specify.
Each of these table transformations are described briefly below:
a) The Group/Summary Command
The Group/Summary command creates a JMP window that contains a summary table. This table summarizes columns from the active data table, called its source table. The summary table has a single row for each level of a grouping variable you specify. When there are several grouping variables the summary table has a row for each combination of levels of all grouping variables.
The summary table has these functions:
• You can add columns of descriptive statistics to the summary table for any numeric column in the source table.
• When you highlight rows in the summary table, the corresponding rows highlight in its source table. If the summary table is in “By-mode,” the highlighted rows in the source table identify subsets. Commands on the Analyze menu recognize subsets identified by a summary table in “By-mode.” A single Analyze menu command produces a separate report window for each subset selected by the summary table.
b) The Subset Command
The Subset command produces a subset consisting of selected rows and columns from the active spreadsheet. To select a column, click in the background area above the column name. To select a row, click the row number in the spreadsheet or highlight the corresponding value in a graphical display. If no rows or columns are selected, then Subset reproduces the entire data table.
c) The Sort Command
Select the Sort command from the Transform menu and complete the sort dialog. Select sort fields from the Columns list and add them to Sort By list with the Add button. You can remove sort fields by selecting them in the Sort By list and clicking the Remove button.
The columns you add to the Sort By list establish the order of precedence for sorting. The first column is the major sort field, and each successive column in the list sorts minor to the previous column. Click the Sort button when you have completed the dialog.
The Sort command creates a new data table. However, you can check the Replace Original Table option to overwrite the original data table with the sorted version.
d) The Stack Command
The Stack command creates a new data table from the active table by ‘stacking’ specified columns into a single new column. To stack columns select the Stack command and complete the stack dialog. Select the columns to be stacked from the Columns selector list and click the Add button. You can enter any name for the new stacked column in the Stacked Column Name box. The type column identifies each row in the new table. The value in the type column is the name of the column in the original table that contained the stacked value. You can enter a name for the new data table in the Output Table box.
e) The Split Command
The Split command creates a new data table by splitting one or more columns to form multiple columns. The new columns correspond to the values of an ID variable you specify. To split columns,
• select each column to be split from the columns list in the split column dialog, and add them to the 'columns to be split' list,
• select a single column from the original data table, whose values become the names for the new split columns and click the ID button. • Optionally, select each column whose values uniquely identify each row in the new table and click the Group button. If you do not specify group variables, the levels of the Col ID variable also define split levels.
f) The Transpose Command
The Transpose command creates a new JMP table that is the transpose of the active data table. The columns of the active table are the rows of the new table, and its rows are the new columns. The columns of the original table must be either all character or all numeric, except columns in a Summary table (see the Group/Summary command) used for trouping. The new table has an additional column whose values are the column names of the original table.
g) The Concatenate Command
When two or more data tables are appended end to end, they are said to be concatenated. The Concatenate command concatenates data tables and creates one column in the new table for each column name in the original tables.
h) The Join Command
The Join command creates a new data table by merging (joining) two open data tables side by side. Tables can be joined
• by row number
• by matching the values in one or more columns that exist in both data tables
• in a Cartesian fashion where all values in a column of one data table are merged with all values in a column of another table.
-- Join by Row Number: The simplest join combines tables by row number. when you match by row number and click Join in the Join dialog, JMP creates the new data table that is the two original tables joined side by side. The new table has all columns from both tables.
If you don’t want all columns from the original data tables, the Select Cols option in the Join dialog allows you to select a subset of columns.
i) The Table Info Command
The Table Info command displays a dialog you can use to make notes about the current data table, examine its attributes, or lock the data table. You can also use the Table Info dialog to create a new data table by simply entering a new table name. The original data table remains as it was when last saved and the new table is not save until you select the Save or Save As command.
j) The Attributes Command
The Attributes command creates a new table, called an Attributes table, from the active data table, called its Source table. An Attributes table has a row for each column in its Source table and a column for each column characteristic
• The Attributes table has a Table Info button in the upper-left corner that accesses the source table's Table Info dialog.
• You can modify the column characteristics of a table by editing values in its corresponding Attributes table row and then choosing the Update Source command in the dollar ($) pop-up at the lower left of the Attributes table. Changing a column's characteristics by editing a row in its Attributes table is the same as changing characteristics in the Column Info dialog for that column.
k) The Design Experiment Command
The Design Experiment command accesses JMP's DOE module, and displays a dialog for you to choose the design type you want. Available design types are two-level, response surface, mixed-level, mixture , and General Factorial. You can also do a d-optimal design search using the D-Optimal personality in the Fit Model dialog. After you choose a design type, specify the number of factors, and select a specify design, JMP builds the table of runs. Also, you have the option of automatically generating a correct analysis for the design.
7. CUTTING, PASTING, AND REPORT WRITING
JMP sports a handy journal feature where you can store interesting results—text and graphs—as you go along. Once placed in a journal, the results can be saved in the format of your choice. Available formats include most of the commonly used word processors.
The Journal can serve as the outline for a report. Just select the Journal command from the Edit menu to create your journal and place the contents of the active window into it. You can append the information from any JMP window to your journal by selecting the window and choosing the Journal command again. You can also copy specific graphs and reports from any analysis window and paste them into the journal or other applications.
And, of course, the Print command works as you would expect it to. You can print analysis windows, data tables, and help windows, at any time during a JMP session.
8. HAVE A GOOD TIME
In the limited space available on this disk, we have tried to give you a sampling of JMP’s most engaging features. But there are many more capabilities which we haven’t even touched on! Don’t be afraid to experiment with the Analysis and Graph platforms. Wherever you see an icon (beneath a plot, beside a plot, or on the window border), click to see what it has to offer. Try sorting or subsetting the demo tables. Use the Join command to reconstruct subsets and the stack command to reshape tables. The files are well documented. When you open a file, check both the Table Info and the Column Info for each column. The Table Info command is in the Tables menu. Also, there the file called SAMPLE DATA readme tells you what each data table illustrates.
