These notes provide an introductory set of exercises to get you started using JMP. They are meant as a starting point only. You are encouraged to explore alternative menu options and data sets. A second objective is to show how JMP tools can not only help you to perform routine statistical analyses, but also, how you can use its visual tools to help understand statistical concepts.
A variety of documentation, including tutorials and books, is build into JMP and can be accessed from the "Help" menu.
JMP 6 is available to Virginia Tech Faculty and Students via an annual licensing agreement with SAS Institute and can be obtained from Information Technology Acquisitions (540-231-5444 or software@vt.edu):
http://network.software.vt.edu
If you receive the error diagnostic "Unable to save menu and toolbar settings to disk" when exiting from JMP, you can correct this problem the next time you open JMP by selecting "Preferences" (from the opening panel or "File" menu), selecting "File locations", and then selecting "JMPCMD file directory". Replace the entry given here by one to which your ID has write access, for example, the one listed under "Preference file directory".
Intel Mac Warning: SAS Institute recommends that JMP 6 not be installed under the "Rosetta" emulation software on Intel Macs as doing so has been found to result in numerical inaccuracies in computed results. for full details, see:
http://www.jmp.com/support/known_issues.shtml
When using JMP, you may find it useful to have several JMP windows open simultaneously. Thus, it is recommended that you close down other applications you are not using prior to starting JMP.
You can start JMP by double clicking on the JMP icon, on a JMP data set (a file ending in ".jmp"), on a JMP report (a file ending in ".jrp" which contains the results of a set of commands), or on a JMP script (file ending in ".jsl which contains a sequence of JMP commands). Once JMP has started, you can save the contents of the currently open data spreadsheet or report by selecting "Save" from the "File" menu; you can save the sequence of commands you have entered to obtain a set of results by selecting "Save Session Script". The results saved in a Session Script can be viewed in any text editor or by selecting "Open" from the "File" menu and then locating the file you wish to open -- be sure the appropriate file type (three letter ending extension) is displayed in the "Files of Type" field or that you have selected "All Files (*.*)".
For our first exercise, we will start JMP by double clicking on a JMP application icon. On MS Windows systems, select "Start", "Programs", "JMP 6", and then the "JMP 6" icon; on Mac OS-X systems, Select "Macintosh HD", "Applications", and then the "JMP" icon. Alternatively, you can use the "Find" or "Search" command to locate the JMP application icon.
To exit from JMP, from the JMP "File" menu, select Quit or press the ¤Q) key sequence.
Note: The "¤ key" notation is used to indicate command key short cuts:
You can open a new data spreadsheet by selecting "File", "New", and then "Data Table". The spreadsheet "Untitled" can be used for entering new data: e.g., from the Rows menu, select "Add Rows" and enter the number of rows to be included in the spreadsheet. Similarly, you can add columns to the spreadsheet with the "Add Cols" option of the Cols menu; however, note that the spreadsheet already contains one column so you are asked to enter the number of additional columns. Click within any marked cell to begin entering data. At a later time, you can add additional rows and columns in a similar fashion. Another way to add a new column to the spreadsheet is to double click in the column heading area to the right of the last defined column of a spreadsheet.
To begin data entry, simply click, or double click, within the desired cell and type in a data value. When you press the "return" key, the data will be entered in the current cell and the cursor will be moved down one row in the same column. If you press the tab key, the data will be entered in the current cell and the cursor will be moved one cell to the right when multiple columns are present, or down one row if only one column is present. If the tab key is pressed when the cursor is located in the rightmost cell of a row, the cursor will move to the first column of the next row; a new row will be created if one does not already exist. Alternatively you can use the arrow keys or mouse to move to a cell and click to begin data entry.
A new spreadsheet can be created at any time by selecting New from the File menu. It is recommended that you periodically save your work as you enter data in a spreadsheet.
As an example, we will create a spreadsheet containing 3 rows and 3 columns. With an "Untitled" spreadsheet displayed on the screen, from the Rows menu, select "Add rows". Enter 3 as the number of rows to add, then select "OK". From the Cols menu, select "Add Multiple Columns", enter 2 as the number of columns to add, and select OK. Enter the data values 1 through 8 into the first eight cells of this spreadsheet. We will not enter anything in the ninth cell to observe how JMP processes missing values in computations.
Note: The "New Column" option under the Cols menu can be selected whenever one is interested in adding one column at a time. When this option is selected, the column definition dialog box appears which allows one to specify the name of the column, data entry validation characteristics, data type, field width, format, and data source. If you use JMP for data entry, the validation fields will prove to be valuable as you can specify a list or range of valid data entries.
The JMP Calculator is accessed from the Column Definition Dialog Box which is opened automatically when "New Column" is selected from the Cols menu. To open the Column Definition Dialog Box at a later time, double select the column name.
Adding values in three columns: From the Cols menu, select "New Column". The Column Definition Dialog Box will appear. Enter Sum for the Column name and select Formula from the "Column Properties" menu. Select "Evaluate Formula". The calculator menu will appear. To facilitate understanding of how the calculator functions, it is suggested that you now position the Calculator and "Untitled 1" windows on your screen so that the new column (Sum) and the calculator are simultaneously visible on the screen. Select "Column 1" in the list of variables and then on the Plus (+) symbol. Select "Column 2", Plus, and then on "Column 3". Then select "evaluate". Observe the computed values displayed for the variable Sum. For each row, except the last, the values in columns 1 though 3 have been added together as the value displayed within this column. Since column 3 of the last row contains a missing value, the third row of the Sum column also contains a missing value.
Averaging values in three columns: From the Cols menu, select
"New Column". The Column Definition Dialog Box will appear. Enter
Average for the Column name and select Formula from the "Column Properties"
menu. Select "Evaluate Formula". The calculator menu will appear. To take the mean of
selected variables in a row, from the Function browser, select
"Statistical". Select the variable "Column1",
press and hold down the shift key, select the variable "Column2", select the
variable "Column3", and then select "evaluate".
Observe that the column
labeled Average now contains the Mean values for the three observations
in rows 1 and 2; in row 3, the mean is based on the two valid
observations for columns 1 and 2.
JMP saves its data in its own proprietary data format and appends the ".jmp" extension to the file name. You can open a JMP data file simply by double clicking on it or by using the "Open" dialog window accessed from the "File" menu. JMP 6 can also read in a variety of other data formats including raw text files, Excel spread sheets, DBase files, and SAS Data Sets. For a full listing, use the pull down menu for "Files of Type" in the "Open" dialog window.
The first data set we will examine will be "Cities.jmp" which is included in the JMP sample data sets.
"C:\Program Files\SAS\JMP6\Support Files English\Sample Data"
"Macintosh HD:Library:Application Support:JMP:Support Files English:Sample Data"
Using the horizontal scroll bar, observe that the "Cities.jmp" data set contains the following variables: city, X, Y, State, Region, pop-m, POP, Max deg.F Jan, OZONE, CO, SO2, NO, PM10, and Lead.
From the Analyze Menu, select "Distribution". To select all variables but city, state, and pop-m for examination, first select Region, by clicking on its name, and then by clicking on "Y, Columns". To select both X and Y, first press the ¤ key, and while it is depressed, selectce on X, selectce on Y, and then click on "Y, Columns". To select the remaining variables, press down on the ¤ key, move the mouse pointer (cursor) to POP, press the mouse button, and while keeping it and the ¤ key depressed, drag the cursor down across the names of the remaining variables. The names of all these variables should now be highlighted; release the mouse button and then the ¤ key. The variable names should remain highlighted. Select "Y, Columns". We have now completed the selection of variables for analysis. Select "OK".
Note: Since the variable City is defined as being of type L (Label), it is automatically used as the label for analyses performed against this data set.
Be patient while JMP computes summary statistics and generates histograms for each of these variables. At this point, you may wish to move and resize the Cities and "Cities: Distribution" display windows. By making the "Cities: Distribution" window large, you can see more of its contents on the screen; by keeping part of both windows visible, you will be able to move between these two windows with a single click of the mouse button.
Note that the data values in the Region variable are nominal values (i.e., letter abbreviations for the state names within which each city is located) and that all other variables are numeric. Thus the display for the Region variable is somewhat different from the other variables: a histogram is displayed in the top half of is display box and a frequency table is presented in the bottom half of the box. For the other variables, you will observe a histogram and box plot (with outlier indicators) in the top portion of the display boxes and the summary quartile and moment statistics in the bottom portions.
By clicking on the blue diamond adjacent to any output heading you can collapse or expand the information displayed for that heading.
By clicking on the red triangle adjacent to a variable name, additional options, including the ability to save the results or a script of the commands used, are displayed. For example, to display a mosaic plot for Region, select the red triangle adjacent to the variable name "Region" in the display heading, and then select "Mosaic Plot". Note that other options are available for the numeric variables; for the variable "CO", select "Quantile Plot".
These graphics provide a visual understanding of the data distribution for each of the variables, but sometimes a formal statistical test is required. To compute a statistic as a test of normality, select the popup menu (red triangle) adjacent to a numeric variable name. For example,onemight ask: "Does the CO levels amongst these cities fits a Normal distribution?" Select the red triangle adjacent to the Variable name "CO", select "Fit Distribution" then select "Normal". Now select the red triangle adjacent to "Fitted Normal" and select "Goodness of Fit"; select this red triangle again and select "Quantile Plot" If necessary, you can use the right scroll bar to move down to see the resultant statistic. Alternatively, you can suppress the displays of the statistics at the bottom of the screen by clicking once on the Quantile and Moments buttons. You can restore one or both of these displays by clicking once on the corresponding button displayed on your screen. Note that the test of normality is performed only for the variable selected. Now perform the normality test and display quantile plots for the variables "POP" and "Max deg.F Jan". Observe the P values of the significance tests and then scroll back up to review the corresponding plots.
You can use colored markers to indicate the location of points within plots. From the Rows menu, select "Color or Mark by Col". A dialog box will appear. Note that an check appears in the block labeled Color, but not in Marker. Select "Region" from the list of variables and then select OK. Observe that each point in the plots appears in a color corresponding to its region (i.e., as depicted in the Mosaic plot for Region); alternatively you could have selected "Mark" instead of "Color" and each region would have shown up with a different symbol rather than color. If there were a second categrical variable with a limited range of values, you could use a different symbol, rather than color, for each level of this variable by selecting "Color or Mark by Col" from the Rows menu and then checking mark and unchecking color. Scroll right and left to observe the colored markers in the normal Quantile plots and the outliers in the box plots. Move the mouse pointer to the Cities window and click. Now scroll down through the list of cites; observe that a colored marker appears in the block in front of the city names. A different color is used for each region and all members of the same region are displayed with the same color.
A variety of tools are available to provide additional fucntionality. These tools can be accessed from the "Tools" menu or by selecting the corresponding icon from the tool bar.
By default, the mouse pointer is used to select items on the screen. In the "Cities:Distribution" window, move the horizontal scroll slider to display the data for the variable POP. The individual points which appear above the box plot are outliers, i.e., points which lie further than 1.5*(interquartile range) from the top of the boxplot. To which city does the uppermost point belong? Move the mouse pointer to this point and press the mouse button. While the button is depressed, the name of the corresponding city will be displayed. You should also observe that a portion of the block in the histogram containing this value is now darker. Scroll right and left. You will note that the portions of blocks in the other histograms corresponding to this point are also darkened. Now click in one of the bars in a histogram and scroll right and left in this window. Observe that the portions of the other histograms corresponding to the block of the selected histogram bar are also darkened. Move the mouse pointer to the Cities window and click. Now scroll down through the list of cites; observe that the blocks in front of the city names corresponding to the selected bar in the histogram are now highlighted.
It is also possible to select observations from a data set and observe where they appear on a plot. For example, while in the Cities window, move the mouse pointer to the column to the left of the column labeled city; observe the portions of the blocks in each hsitogramwhich are now highlighted. Now press the ¤ key, click the mouse button, and while both buttons are depressed, drag the pointer vertically across 10 to 15 rows, release the mouse button, and then release the ¤ key. You can also use the ¤ key to add individual cities to those already selected, e.g., press down on the ¤ key, and while it is depressed, click in the block to the left of one or two additional city names. The leading column in front of these cities should now be highlighted. Click in the "Cities: Distribution" window. Observe that the locations of these cities within the histograms are now highlighted.
The Hand tool can be used to modify the space and center point location of the bars in the histogram. Move the mouse pointer to the Tools menu and select the Hand ("Grabber") tool. Now when you move the mouse pointer into a graphics window, it takes on a hand shape. Move it into one of the histogram windows and press down on the mouse button. As you move the hand tool vertically, observe how the center location of the histogram bars change. Move the pointer to the right and observe that the number of bars increases; move it to the left and note that the number of bars decreases.
Note: Although we have changed the appearance of the plots using the Hand tool, we have made no changes to the data upon which these plots are based.
Select the Crosshairs tool, from the Tools menu. When the mouse pointer is now moved into a graphics window, it changes to a plus symbol. Press the mouse button and the horizontal and vertical values of the point will be displayed.
The Scissors tool can be used to select parts of a plot to be copied (from the Edit Menu) into another application -- first use the "selection Tool" (select "Tools" and then "Selection") to activate this tool and then use it to select the component or components you would like to copy to another application. The Magnifying glass tool can be used to increase the plot by about 25% for each mouse click. Depress the Option key and click anywhere within the plot to restore it to its original size. The Annotate tool can be used to add text to the plot. The Help (?) tool is used to obtain help regarding the current application.
To restore the normal (default) mouse cursor, from the Tools menu, select the Arrow tool.
You can adjust the size of a plot by positioning the arrow pointer on one of the edges or corners of a plot, pressing it down, dragging it to a new position, and them releasing the mouse button.
Cross Tabulation: To perform this analysis, we will use another data set. From the File menu, select Open (¤O). Select "CAR POLL". Now select "Fit Y by X" from the Analyze menu. Select "Country" and "X, Factor". Select "Size" and "Y, Response". Select "OK" and observe the results. Note the display of the two-way mosaic plot and two-way contingency table followed by test statistics.
From the "Windows" select the Cities data set. Earlier we had used this data set to look at one of the choices available from the Analyze menu: "Distribution of Y" which is used to obtain univariate statistics. We will now look at multivariate choices available from this menu.
Linear Regression: we will now examine the possible relationship of CO (Carbon Monoxide) to population size in the Cities data set. From the Analyze menu, select "Fit Y by X" and a dialog box will appear. To specify population as the independent variable, select the variable name "POP" and then on "X, Factor". Now select the variable name "CO", and then on "Y, Response" to specify the dependent variable. Select "OK". A plot of POP by Y will appear. From the Fitting popup menu in the lower left hand corner of this window, select "Fit Line". An Analysis of Variance table and regression parameter estimates are displayed.
Note: If you had assigned colored markers to any of the cities, these markers will continue to appear on output plot displays.
Multiple Regression: Do the X and Y coordinates (Longitude and Latitude) or Maximum January Fahrenheit temperatures contribute to the above regression model? To run a multiple regression, from the Analyze menu, select "Fit Model". To include CO as the dependent variable in the model, select POP and then on "Y, Response". From the list of variables, select POP and "X, Factor", X and "X, Factor", Y and "X, Factor", and "Max deg.F Jan" and "X, Factor" to include these as independent variables in the model to be tested. Then select "OK" and observe the displayed results. Scroll right to see the Leverage plots generated for this model. For an explanation of the Leverage plots, select the question mark tool from the Tools menu and then click inside one of the Leverage plots. Restore the normal cursor by selecting the arrow tool from the Tools menu. You can use the Check popup menu in the lower left to obtain a residual plot (Plot Residual).
Again select "Fit Y by X" from the Analyze menu. Observe the table displayed in the lower left of the dialog box which appears. This diagram indicates the type of analysis which will be performed based upon the characteristics of the X and Y variables. As we have seen above, if both X and Y are continuous variables, a regression analysis is performed.
Note: As we have seen above, a Cross Tabulation (Contingency Table) analysis would be performed if both were either nominal or ordinal. Logistic regression is performed when the "Y, Response" variable is nominal.
Correlations: How are variables in the data set correlated with one another? From the Analyze menu, select "Multivariate Methods", and then select "Multivariate". A dialog box will be displayed. Select all of the variables in the data set, and after they have been highlighted select "Y, Columns". Note that a pop up window warns you that the non-numeric variables "City", "State", and "Region" had been selected; only numeric variables are included in the list of variables to be used for the analysis. Select "OK". A correlation matrix and matrix of scatterplots will be displayed. Use the vertical and horizontal scroll bars on this window to move around and observe the degree of correlation between various variables in the data set.
A one-way Analysis of Variance is performed if the independent variable "X, Factor" is a nominal variable and the dependent variable "Y, Response" is a continuous variable. For example, select "Fit Y by X" from the analyze menu. Select Region and then select "X, Factor". Select "Max deg.F Jan" and then select "Y, Response". Select "OK". From the Analysis popup menu in the lower left hand corner of the screen, select "Means, Anova/t-Test". Observe the results. From the Analysis menu, select "Compare all Pairs" and observe the comparison circle plot that is generated. Now from the Analysis popup menu, select "Unequal variances" and scroll down in the current window to read the results.
Here we see a warning message which suggests that we use caution due to small sample sizes. Note that the data set includes two or fewer observations for the regions C and TX. What would happen if we were to exclude these observations from the analysis? From the Windows menu, select "Cities". Press down on the ¤ key, and while it is depressed, click in the area to the left of the cities Galveston, Houston, Los Angeles, and San Francisco. Four cities should now be highlighted. From the Rows menu, select Exclude. A circle with a slash through it should appear in the highlighted area adjacent to each of these names indicating that they will be excluded from the analysis. Repeat the procedure for performing the Analysis of Variance and again use the Analysis popup menu to select "Unequal variances".
Factorial Analyses are performed using the "Fit model" choice of the Analyze menu. To illustrate a three-way analysis of variance, we will use a new data set. From the File menu, select Open, select POPCORN from "Sample Data", and then select "OK". This data set describes popcorn yield based on three factors:
Which of these factors, or their interactions, contribute most to the value of yield? From the analyze menu, select "Fit Model". To define the dependent variable, select "yield" and then on "Y". To select the independent variables, press down on the mouse button, and while it is depressed, drag across the variable names "popcorn", "oil amt", and "batch". For a main effects only model, in the "Construct Model Effects" pane, select "Add" and then select "Run Model". Scroll right in the "POPCORN: Model Fit" window and observe the Leverage plots and computed statistics.
For a full factorial model, select the three independent variables as above and then in the "Construct Model Effects" pane, from the "Macros" popup menu, select "Full Factorial". Observe that the selected variables and their interactions are now displayed. Select "Run Model". Scroll down in the "POPCORN: Model Fit" window and observe the plots and computed statistics.
The "Cross" menu button provides flexibility to include only specific interactions of interest. Say we wished to include all three factors, but only the "popcorn*batch" interaction. Select the variables "popcorn", "oil amt", and "batch" as above for a main effects only model. Then select the variables "popcorn" and "batch" followed by selecting "Cross" in the "Construct Model Effects" pane. You should now see the three variables followed by the "popcorn x batch" interaction displayed. If desired, additional interactions can be specified in a similar fashion. In some cases, it may be easier to specify a model using a choice from "Effect Macros" and then use the "Remove" button after highlighting interactions which are not needed in the analysis. Now select "Run Model" and observe the results.
JMP can also be used to design experiments using the DOE menu. Here you will see choices of a variety of designs including: "2-Level Design", "Response Surface", "Mixed-level Design", "Mixture Design", "General Factorial", and "D-Optimal Design". Once the design has been created, it can be saved to disk and then retrieved for analysis after the data has been collected.
We will use another data set to examine the capabilities of JMP's graphic tools. Instead of using the Open command from within JMP, we will select the desired data file and thus invoke JMP to load this file into JMP memory. Open the folder containing the JMP Sample files (see ACCESSING EXTERNAL DATA FILES) and select the file "Iris.jmp". Alteratively you can use your systems "Find" or "Search" command to locate the file "Iris.jmp" and then double click on the corresponding icon.
The spreadsheet that now appears contains Fisher's classic data for sepal and petal measurements of three iris species. Note that there are 150 observations in this data set. The first 50 observations correspond to the species setosa and use a red square as a marker, the second 50 observations belong to the species versicolor and use a green square as a marker, and the last 50 observations correspond to the species virginica and use a blue square as a marker.
From the Graph menu, select "Overlay Plots". A dialog box will appear. Select "Petal width" and then on "X, Factor". Select "Petal length" and then on "Y, Response". Select "OK". A plot showing each of the data values for these variables is displayed.
As another example of using overlay plots, from the Graph menu, select "Overlay Plots". A dialog box will appear. Select "Petal width" and then on "X, Factor". Press the ¤ key and while it is depressed, drag the mouse over "Sepal length", "Sepal width", and "Petal length"; then select "Y, Response". Select "OK". A plot showing these three variables plotted against "Petal width" is displayed. To obtain separate plots for these three variables, select "Overlay" from the "Red Triangle" pulldown menu in the heading for this plot.
From the Graph menu, select "Spin plot". A dialog box will appear. Press the ¤ key and while it is depressed, drag the mouse pointer across the variable names "Sepal width", "Petal length", and "Petal width". Select "OK". A window containing a 3D plot for these variables is displayed. Clicking on the Arrow icon boxes in the upper left hand corner of this window is used to turn the display and thus change the viewing perspective; once a spin direction has been selected, the plot will spin continuously when the pointer is in the plot area and the the ¤ key and mouse button are depressed. If you would like to change to a white plot background, select the red triangle amongst the spin control buttons and then select "White Background". The Hand tool can also be used to turn the plot. When you are finished viewing the spin plot using the Hand tool, select the arrow tool from the Tools menu.
Select "Contour plot" from the Graph menu. A dialog box appears. Select "Sepal width" and then "X". Select "Petal width" and then "X". Select "Petal length", select "Y", and then select "OK". A contour plot showing the relationships of these three variables is displayed. From the Check menu in the bottom left hand corner of the screen, select Contours. Check the box "Fill Areas", then select "OK". Observe the new display.
From the Graph menu, select "Ternary plot". Again select the variable names "Sepal width", "Petal length", and "Petal width". Select "Y, Plotting", and then select "OK". Observe the new display.
Occasionally while running JMP, you may receive a message indicating that you are running low on memory. Typically this problem can be resolved by closing some of the open windows. It is good practice to Save (¤S) your work to files when you complete each analysis and close the associated window, but it is often desirable to keep a window open to compare its contents with the results of a subsequent analysis. The Windows menu provides a mechanism for keeping track of and switching between open windows. Each of the currently open windows is listed in this drop-down You can bring a window to the foreground by selecting its name from this list.
Let's now close all of the open windows. You could click in the
close box of each open window, but this is tedious if many windows are open.
Instead, you can close all windows of a given type simultaneously by using the
"Close All" option under the Windows menu.
What criteria would you specify for a healthy breakfast cereal? The JMP cereal data set ("cerea.jmp") contains data on calories, protein, fat, sodium, fiber, complex carbohydrate, sugar, potassium, and degree of enrichment for 77 popular breakfast cereals. On the basis of on your own selection criteria, use the JMP tools to identify the cereals in this data set whose consumption would contribute to your definition of a healthy life style using the provided varaibles. What patterns can you find among these variables?
Examine other data files available via the Open command under the File menu. When you are done, from the File menu, select Quit (¤Q).
Open your own data JMP and begin using the JMP tools to gain new insights into the information contained within it.