Moving from SPSS to JMP : A Transition Guide

Transcription

1 WHITE PAPER Moving from SPSS to JMP : A Transition Guide Dr. Jason Brinkley, Department of Biostatistics, East Carolina University

2 Table of Contents Introduction... 1 Example... 2 Importing and Cleaning Data... 2 Visualization... 6 Descriptives Descriptive Statistics Custom Tables Correlation Inference Two-Sample t-test Crosstabs/Contingency Tables Linear Regression Data Manipulation Creating New Variables File Splitting Saving and Reproducing Output Other Features Conclusion Acknowledgements i

3 Introduction There is an increasing multitude of software options for processing and analyzing data. The way each piece of software handles, processes and outputs analysis can be quite different. This can make transitions from one software program to another quite difficult. In fact, it can be just as difficult to go to another version of the same software program. Sometimes novice and infrequent users struggle with changes in the same program or transitions in analytic programs. Many such users remember the way to perform specific analysis or how to interpret output only by the means they are used to performing such functions in their originally chosen software. Thus when a switch is to be made to another software, there can be many transitional issues. SPSS is a well-known and highly utilized program for the statistical analysis of data. It has a rich history in both academia and industry, and is the standard in particular disciplines. Originally developed in the pre-windows graphical interface era, SPSS found a niche early on by being among the first to switch to the graphical user interface format (GUI), which is more commonly referred to as point and click. SPSS developed a graphical front end that allowed users to specify what analysis and options they would like the software to perform without the need to write code. In doing so they delivered analytics to the masses, those unable or unwilling to work in the programming language environment. JMP is a GUI-based analytics software developed by SAS. Originally intended exclusively for Mac users, the JMP division has a more than 20-year history of delivering high quality analytics in the point-and-click manner to users in many different areas. It is important for those transitioning from SPSS to JMP to realize that there are some fundamental differences in how the software operates, which can boil down to a matter of perspective. The SPSS point-and-click interface is a vehicle for developing and implementing SPSS code. Users select which analysis they want to perform and the software writes and submits the subsequent code. The appropriate code is saved to the output file from which users can save and potentially resubmit various routines. New users to JMP need to know that JMP is designed so that users have a dynamic link between their data and analysis. This link leads to a different perspective for data analysis, one which we will explore in this guide. SPSS requires users to specify which analysis they would like to perform prior to seeing output, while JMP prefers users to specify a general direction first and then customize interactively. In addition, in many instances JMP prefers to combine sets of analysis that are all under one general idea and put them together in one platform for convenient access. Perhaps the best way to illustrate the differences and how to transition from SPSS to JMP is via an example. The example for this guide is independent of the software under consideration, so the background and data sets come from alternate sources. In order to be as up-to-date as possible, this transition guide uses the most recent available versions of SPSS (version 19) and JMP (version 9). 1

4 Example Students in an introductory statistics class, taught by Professor John Eccleston and Dr. Richard Wilson at the University of Queensland, participated in a simple experiment. The students took their own pulse rate and then were asked to flip a coin. If the coin came up heads, they were to run in place for one minute. Otherwise they sat for one minute. Then everyone took their pulse again. The pulse rates and other physiological and lifestyle information are given in the data. This is a commonly used data set for teaching statistics and can be found online at the Australasian Data and Story Library via the website Importing and Cleaning Data We start by opening the data; the tab-delimited text file available at the above website can be modified into many formats but can also be directly imported into either software. Let s start with opening the file and viewing data in SPSS. SPSS Upon opening SPSS, users can open the text file using the options under the File menu. After a series of steps where the software checks for things like delimiters, menus, etc., you eventually end up importing the following file: A quick glance at the spreadsheet reveals 11 variables and 110 observations. We see data on height and weight (measured in cm and kg respectively), age, gender (male = 1), smoking and alcohol use (yes = 1), exercise frequency (1 = high, 2=moderate, 3=low), ran or sat status (1=ran), baseline and follow-up pulse, and year the class met. 2

5 SPSS has two methods of looking at each data set, the Data View and Variable View options, which can be seen via tabs on the bottom of each data set. When we open the data in Variable View, we see what type of data each variable is and can modify that information by adding labels and changing format. There are already a handful of things that need to be taken care of to clean this data and get it ready for analysis. Pulse 1 and Pulse 2 variables were imported in nominal format, and the original data has an observation with no measurement for observation 76 (only a listing of NA ). In order to perform proper analysis on this data, we need to inform the software that this is a missing value. We can start by clearing the NA and then changing the variables format. So under Variable View we go from this: To this: In doing so we also change the format of Year to ordinal in order to ensure that it is of the appropriate type for analysis. Next, we may want to label categorical variables for other analysis. This is also done in the variable view table using the Values option. 3

6 So we see that the basic idea of presenting data in SPSS is to provide two viewing options that show the data in terms of the individual observations, as well as an overview of each variable s type. JMP Switching over to JMP, we will open the same data and do the same manipulations. Let s begin with what happens when starting the program. When JMP starts the first thing users receive is a Home Window screen, which manages different windows for data, log and output. Recently used files are available over to the left side so that users can conveniently reopen files they were using in a previous session. Another feature of JMP is the JMP Starter window, which is found under the View menu. The JMP Starter is convenient for new users by organizing commands for analysis into categories and providing summaries for each data and analytic option. We can open the data for this project by clicking on the Open Data Table command in the File category. You can also open data directly from the File menu. 4

7 Like SPSS, JMP can open a variety of files including Excel spreadsheets, SAS data sets, SPSS documents and text files. We can import the data from our example and see that it looks similar to what we have already seen in SPSS. Unlike SPSS, JMP does not switch between multiple views in the data. Instead, there are table panels along the left side that do similar things. The middle left panel displays the column information and shows each of the 11 variables and the data type. There are modeling-type icons for each data type: blue for continuous variables, red for nominal data and green for ordinal data. Note that Year was changed to ordinal (simply click on the icon) to make it appropriate for analysis. The bottom left panel summarizes the rows and will show additional information that we will discuss later. As in the SPSS file, there is a problem with the format of variables Pulse 1 and Pulse 2. After removing the codes for missing data, we still need to tell JMP that this variable is continuous. By right-clicking on the Pulse 1 variable in the column panel, we can go to Column Info and change the data type. The Column Info pop-up window also lists a set of column properties that can be used to add labels to the data, so we can pull the column info for gender and add the labels 1 for males and 2 for females. When columns are given properties, JMP signals these changes with different symbols in front of the column names of interest. Here we see that asterisk symbols have been created for both Gender and Smoking Status variables. 5

8 Visualization Visualization is a powerful tool for analyzing and understanding data. The ability to easily make graphs via statistical software is one of the principal reasons why analytic software is so popular. We will confine our discussion to three simple tools: bar charts, histograms and scatter plots. SPSS Suppose we want to limit the discussion to the variables Height, Weight and Gender to visualize the distribution of each and perhaps see all three on one graph. We start with forming a bar chart of Gender. One quick and easy way to do this is with the Legacy Dialogues under the Graph menu. For this data we see that slightly more than half of the data points are male. Turning to Height and Weight, we can visualize this data in a histogram, which can also be found in the Legacy Dialogues. 6

9 In addition to the histograms, we also get some simple statistics for each variable. We can double-click on each graph in the output and customize as we see fit. Examining the distribution of heights and weights, we see two observations whose heights are much shorter than the rest of the data. For more complicated graphs, we use the Chart Builder option under the Graph menu. Using Chart Builder, we can build a scatter plot of height versus weight that is color-coded by gender. The Chart Builder dialog box is useful because it s easy to use and can be customized. Simply click on the type of graph you want to make and drag it into the preview box. From there, click and drag the variables you want to use on the x- and y-axis and overlay color. Once the graph is set we can view our results. In viewing this scatter plot, several points about the data can be made. First, the unusual observations for height are still an issue, the lowest value of height in particular. That data point certainly looks out of place with respect to the rest of the data. Besides this insight, we also see a linear relationship between height and weight and that the males tend to be taller and heavier than the females. Given that the unusual point is male adds to the suspicion of a possible outlier. 7

10 JMP Before we look at data visualizations in JMP, recall from the introduction that it was stated that JMP was designed to be interactive software. The majority of this interactivity comes in the form of little red triangle icons in various places on the data and output that enables the user to explore additional analysis. We start again with a basic bar chart. Users can access graphing features either by the Graph menu in the JMP Starter window or the under the Graph menu item from the top menu bar. To make a bar chart we select the Chart platform. The Chart platform lets users make a wide variety of graphs, such as pie charts, bar charts and line graphs. In the launch window users select Gender from the Select Columns list and then click Categories, X and Levels to load the variable. It is important to note that if the data is shown as a summary count instead of individual rows, we can still form a bar chart. Note that the Chart platform in JMP makes bar charts for raw data as well as summaries. In SPSS, users have to weight summarized data by counts prior to creating such a graph. In JMP, users who have data in the form of summarized counts simply pick the N option from the Statistics menu to tell JMP that the data is summarized counts rather than in raw format. 8

11 The standard report is a simple bar for each group along with a red triangle icon to manipulate the report. Some options for manipulation include changing the chart from vertical to horizontal format, making pie graphs, adding percentages and manipulating the axis scale. We begin to see the difference in the way JMP operates. Rather than changing options prior to seeing the graph, users are allowed to look at the graph and make changes that are instantly reflected in the report. An important option is the Recall button that reproduces the last set of analysis used in that platform for that session. So if you accidently close the wrong report, it can easily be retrieved. Moving on to histograms, we find a shortcut button when Graph is selected on the JMP Starter. Starting with Height and Weight variables we can obtain simple and straightforward output. As with the chart report, there are red triangles that allow certain customizations such as showing the percentage of the data that is in each bar. Users can also make histograms using the Distribution platform from the Analyze menu, which we will discuss later. 9

12 Again, we see two unique individuals who have heights much smaller than the other individuals. We list them as possible outliers, but here is where the differences between SPSS and JMP truly begin. JMP allows users to dynamically interact between the data and reports. For example, users can click on individual bars on a histogram to highlight a certain subset of the data. Here we are calling attention to the potential outliers. There is similar shading in the weight histogram, indicating the weight values for the selected heights. We see that one observation with a low height does indeed have the lowest observed weight, while another has a weight value in the middle of the data. Returning to the data table, the selected rows show up as highlights in the data table. This gives users the option to quickly inspect the rows to determine exactly which observations are causing the controversy. If the data needs to be checked for consistency, then having the data and the output dynamically linked makes it easy to go back and inspect individual observations. There are many ways to clear data selections, but one of the easiest is just to double-click on Selected in the Rows panel of the data table (likewise double-clicking on All Rows selects all the data). Moving on to the JMP Graph Builder platform (located under the Graph menu), JMP creates an environment for building a graph similar in spirit to the Chart Builder in SPSS. However, instead of preselecting a graph type and then setting the variables, JMP wants users to set the variable roles first. To create the same graph as earlier, we can click and drag Height to the x-axis, Weight to the y-axis, and Gender to the Color zone. Graph Builder gives you many other options to overlay and stratify data. Different graph types and options can be obtained by right-clicking on the background of the graph (such as adding or removing the Smoother function). New in JMP 9 is the Shape zone, which can be used to quickly create map shapes such as US state or county maps. 10

13 After the graph is appropriately customized, then it can be copied and pasted into many outside programs such as Microsoft Word and PowerPoint. All JMP reports can be copied in such fashion, but by default the menus for saving, editing and copying in a JMP Report window are hidden. New users need to be aware that starting with JMP 9, one must click on a thin strip at the top of the report to get the hidden menus to appear. Users can change this option by going to Preferences under the File menu and looking for Window Specifics. From Graph Builder, we can begin to see exactly how dynamic linkage separates working with JMP from working with SPSS. Users can click on individual points in these graphs and select them for dynamic analysis. In SPSS we observed the possible outliers in the graph, but JMP takes it a step further and allows users to highlight those individual points and see them in the data and/or other report. In addition, one can exclude certain data points from analysis without having to delete them. Excluded data stays within the data table, but is not considered in analysis, thus making it easy to do analysis with and without potential outliers. There is also the Hide option if one wants to remove individual points from visualizations, but not analysis. 11

14 Descriptives Descriptive statistics are at the heart of any good statistical analysis and when performed properly, provide real insight beyond what is seen in visualizations. This section will focus on univariate descriptions of continuous data and creating custom tables. Assuming we want to continue to study the variables Height and Weight, we will leave in the potential outliers and consider that data because it is still unclear whether the potential outliers should be removed from analysis. Descriptive Statistics SPSS Many SPSS users know that the Frequencies dialog box can be a one-stop option for analysis of many different variables simultaneously. While there is a Descriptives dialog box, the Frequencies dialog box is nice because one can request descriptive statistics and graphs in one place. Also, users can put several different types of data into this option and SPSS will produce the appropriate graphs and statistics for each type of variable. The options selected here give the standard univariate report, where we see the typical measures of center and spread that are discussed in basic statistics courses. But frequency tables, histograms, bar charts and many other useful descriptive outputs are also available. 12

15 JMP Similar to the Frequencies option in SPSS, JMP has the Distribution platform to obtain simple descriptive output. The Distribution is found under the Analyze menu. We start by selecting Height, Weight and Gender and continue the tradition of manipulating output after we have selected the variables we want to study. This is a standard distribution report. By default, quantiles and moments are produced for continuous variables as well as a 95 percent confidence interval for the mean. Frequency tables are produced for categorical variables, and the male subgroup has been selected to highlight the interactivity between different portions of the same report. 13

16 By clicking on the red triangle, users can customize report display, obtain more descriptive measures or fit different distributions to the data. Add-ons include box plots, normal quantile plots, CDF plots and distribution fits. Being able to work with the output interactively allows the users to fit many different theoretical distributions to the data (i.e., normal, exponential, Poisson, binomial, normal mixtures, etc.) and assess their relative fit to the data. New to JMP 9 is the option to save distribution reports as Adobe flash files that can be viewed and worked with interactively without the use of JMP software. These files make for good demos to students or can be worked into reports so clients can discuss results with consultants. Custom Tables SPSS In many cases, specific output is necessary; like in our example, where the visuals showed differences in height and weight by gender. SPSS also has options to create custom tables for specific output. We can click and drag the variables into the proper locations and specify what descriptive output we want to see. An example of this output is seen below. 14

17 JMP JMP makes custom tables via the Tabulate platform under the Tables menu. The Tabulate platform can build a table interactively, similar in spirit to the Graph Builder. Users see computations and output as they begin customizing their table. In addition, the interactive table can be turned off in favor of a traditional dialog box if the user wants to make a very specific output. Correlation SPSS Using the Bivariate dialog box inside Correlate under the Analyze menu, we can obtain sample correlation estimates between height and weight. Note that there is significant positive correlation between the two variables. 15

18 JMP JMP groups sets of analyses into one platform that allows users to do many different things (all somewhat related) in one common set of reports. For example, the easiest option for finding Pearson correlation coefficients is to use the Multivariate platform under the Multivariate Methods of the Analyze menu. The Multivariate platform is much more than a simple vehicle for finding correlations; however the basic report shown here delivers both a Pearson correlation estimate and a scatter plot matrix. By clicking on the red triangle, the histograms can be added along the diagonal so that both the univariate and bivariate distributions can be quickly visualized. The Multivariate platform also allows users to test correlations, find alternative measures of correlation (i.e., Spearman and Kendall), confidence intervals, partial correlations and visuals such as Color Maps. In addition, correlationbased measurements such as principal components analysis, outlier analysis and item reliability are all available in the Multivariate platform. 16

19 Inference We now move on from descriptive statistics to general inference between two variables. The discussion here will focus on two-sample t-tests, chi-square tests for tables, and regression. Both SPSS and JMP have many other great analytic features for descriptive and inferential statistics, but these three are some of the most commonly used and generally understood methods. Two-Sample t-test SPSS Suppose we want to test whether the heights significantly differ between men and women. The standard independent samples t-test is a popular option in SPSS, which is found under the Compare Means menu. Here we can specify one or more continuous measurements and determine if the mean value of that measurement significantly differs between two groups. We see that there is a significant difference between the heights of the males and females within this data set. In SPSS, both the equal variance and unequal variance tests are calculated and users can rely on a test for variances to determine which test is the most appropriate. 17

20 JMP While there is a button in the JMP Starter menu for t-tests, users should be aware that the t-test and one-way ANOVA analyses all go into the same platform. In fact, JMP combines all two-variable inferences into one Fit Y by X platform. Using the Fit Y by X platform, users can make inferences on bivariate comparisons for any type of data. The grid in the bottom left corner of the launch window lists different analysis types. Users choose which x and y variables they want to study and JMP chooses the appropriate method based on what type of data is being used. For now, we want to examine the differences between heights based on gender. In the report, we have simple descriptives along with a t-test that assumes unequal variances (found by selecting t-test from the red triangle icon). Users may add graphics (such as box plots) and additional tests looking at the variances, or nonparametric tests more appropriate for skewed data. 18

21 Again, we see an important distinction in JMP with the combining of many similar tests into one common platform. Users can start with a t-test and look at graphics and summary statistics to determine if the assumptions of a t-test are met. If they are not, then the user may add on an alternate test that is more appropriate for the data they are considering. Crosstabs/Contingency Tables SPSS Creating a contingency table and performing a chi-square test of categorical data is an important component to many analyses. Examination of table data is one area where SPSS has always focused, and users can find many options for analysis under the Crosstabs listing in the Descriptive Statistics menu. The Crosstabs menus allow users to ask for a number of different sets of analysis based on what type of comparisons the user wants to make. 19

22 From this example we see that even though the rate of smokers among alcohol users is more than twice the rate of smokers among non-alcohol users (13.2 percent versus 4.8 percent), there is not enough evidence to conclude that there is significant association between alcohol use and smoking status (chi-square p-value = 0.15). SPSS does give us a warning that the expected counts are small and provides a Fisher s exact test of the difference among samples with small cell counts. Users interested in relative risk and odds ratio estimates can check the risk option in the statistics box to gain those estimates. JMP Users can access the Contingency Analysis options in JMP via the Fit Y by X platform. Simply specifying categorical variables for both x and y will lead users to contingency analysis. We can specify the same table as we did in SPSS to look at the relationship between alcohol and smoking status. The JMP report gives table information, a chisquare test and Fisher s exact test (2 x 2 table only, same as SPSS). Users can then add on additional summaries and tests such as relative risk, odds ratios and measures of association (for items such as Kendall s tau). The image to the left is considered a standard report, but we can also see the list of options that are available via the red triangle icon. An interesting aspect to JMP is that the software tries to help users in areas where they may be unfamiliar. By selecting the odds ratio option and simply hovering the mouse over that area, an explanation box appears that indicates what kind of analysis this option will add on. For users having trouble interpreting statistics, p-values or other output, this option is available by simply circling or hovering over an area with the mouse. 20

23 Linear Regression SPSS Simple linear regression can be performed using the Linear Regression option under the Regression menu. Suppose we want to use height as a predictor of baseline pulse rate. The Linear Regression option in SPSS is a powerful tool with many different options for model-based estimates, measures of fit, diagnostics, predicted value output and visual plots to assess fit. Users can incorporate many potential predictors and look at both simple linear regression and multiple regression in the same set of menus. Without using additional options, we obtain standard output with R-square, ANOVA table and a table for regression estimates. We see a significant relationship between height and baseline pulse and that as height increases, the baseline pulse of these individuals tends to decline. It is important to note that the linear option does not allow string variables to be used as potential predictors. In fact there are many different regression options in SPSS, depending on what type of model needs to be fit. The linear option is limited to numeric covariates only, and users for multiple regression with both continuous and categorical covariates need to refer to the General Linear Model platform. 21

24 JMP Users have two options for regression models in JMP: the Fit Y by X platform or the Fit Model platform both under the Analyze menu. Simple bivariate fits go along the Fit Y by X platform, so let s start there. 22

25 The Fit Y by X platform can again be employed to look at the relationship between two different variables. Here we have listed Height as X and Pulse 1 as Y. It is important to note that the Fit Y by X platform ONLY looks at bivariate relationships. It is possible to select more than one candidate for both the X and Y roles. In such cases, output consists of multiple bivariate fits with each output displaying the appropriate measures for each set of variables. The Bivariate Fit report starts with a simple scatter plot and the user must assess exactly what type of bivariate fit they would like to make. Here we see a linear fit that models a simple linear regression line to the data. A standard report includes the same basic information as found in SPSS: R-square, ANOVA and parameter estimates with hypothesis tests for the intercept and slope parameters. However, JMP users can interact with this output to look at multiple fits simultaneously. For example, one can choose a simple mean, straight line and quadratic fits to the data and compare the fits by either R-square or looking at the parameter estimates. Other fits include density ellipses and smoothing splines, from which comparisons between fits can be made and estimated values can be output to the original data for further study. Under each fit, users can specify diagnostic measurements and plots to assess whether the assumptions for using each particular model are satisfied. 23

26 While more complicated regression is a difficult matter in SPSS, where users must distinguish what type of regression is necessary from many different options, JMP puts many different types of regression models all under the Fit Model platform. In the example shown here, we are fitting a multiple regression model that predicts baseline pulse via a combination of height, weight and the categorical variable, gender (no need to separate categorical and continuous covariates). Had the response been categorical, then either the nominal or ordinal logistic model would have been the default option for type of analysis. A generalized linear model is available for Poisson or binomial type fits, proportional hazards models for survival data, and MANOVA for repeated measures analysis. For model selection, the stepwise regression option is available and works with any response type, continuous or categorical. Data Manipulation In addition to high quality graphics and analytics, both SPSS and JMP are excellent for manipulating data. SPSS and JMP provide a number of different ways to organize and manipulate the data. Users can sort data, sort variables, merge files, recode and run many other manipulations. Listed below is a side-by-side comparison illustrating a number of the data manipulation features in SPSS and JMP. New JMP users should know that since the data file and the output are dynamically linked, there are occasions where output needs to be closed before the data is manipulated. When sorting or manipulating data, JMP will sometimes default to creating a new data table that has the necessary manipulation; new users should specify that the manipulated file should replace the current data file. 24

27 SPSS JMP 25

28 Creating New Variables SPSS Suppose that we want to create a new variable that is the difference between baseline and follow-up pulse rate. We can use the Compute Variable option to define a new variable (call it Pulse_Diff) and attach a formulaic calculation to that variable. SPSS has a number of different functions for aggregating and transforming data, which are all grouped by function type. After filling in the numeric expression, SPSS creates a new variable of the specified name and populates it with the computation from the numeric expression. JMP JMP creates new variables directly in the data table. Suppose we want to create the same new variable Pulse_Diff in JMP. We start by double-clicking on the unused column at the end of data table to make a new column. Then we double-click on the column name to bring up the Column Info window where we give the column a name, and click on Column Properties and choose Formula to bring up the Edit Formula window. Now, we can edit the formula to create the numeric expression we used in SPSS. Note that JMP also has a wide array of functions that can be applied to manipulate and transform our existing data. 26

29 File Splitting SPSS Suppose we want to look at different output for men versus women (given that men are taller and height is related to baseline pulse). We can use the Split File option to tell SPSS that further output should always be stratified by a particular variable. Users have the option of creating new files for just men and just women, but the splitting option is a unique way to indicate to SPSS that subsequent output should be separated between women and men. Consider the following descriptives output that was automatically stratified by gender. 27

30 SPSS users who want to return to a nonstratified analysis simply return to the Split File option and choose Analyze all cases, do not create groups to turn the stratification off. JMP Splitting a file may not be necessary in JMP since all analysis platforms feature a By option to stratify output by a particular variable. So if we want stratified descriptive statistics then we use the By statement in the Distribution platform as seen below. 28

31 Saving and Reproducing Output Manipulating data and producing great output and visuals are only great if there is a mechanism to save and reproduce them. It doesn t seem obvious but JMP is also fueled by a scripting language, which works mostly behind the scenes for the casual or new users. Advanced users many times find that coding produces quicker results for repetitive tasks. Let s discuss how to save and reproduce output in SPSS and JMP. SPSS When users save an SPSS data file in SPSS format (.sav file) the data file maintains all manipulations, transformations and labels. In order to retain output, we save as a separate file (.spv format). Since the point-and-click interface for SPSS is really a mechanism for code generation, code precedes all output in the output file. For instance, if I want to look at the descriptive statistics for height and weight, the following code could be generated in the output from the point-and-click interface: Users can easily reproduce the exact output by simply retaining and reusing the code that is located in the output file. Therefore, saving both the data and output in SPSS format makes it possible for users to continue to work on the file by reopening them in SPSS and running saved syntax. JMP JMP software takes a different approach to saving and reproducing output. When you save data as a JMP data file (.jmp file) the data retains manipulations, transformations, labels AND script saved from analysis. Let s look at an example: Suppose we want to save the distribution analysis of height and weight information for later use. In all JMP reports, users can find an option for saving the code (referred to as JMP script) to the data table. There are other options to save JMP script, but saving to the JMP data table is one option. When users save the data in JMP format, all scripts saved to the data table are saved as well. Users can edit the name of the saved script and reproduce the output with the red triangle located by the saved script s name in the data table. In addition, JMP users have one file that contains both the data information and analysis related to a specific project. JMP data files can be shared among many JMP users and one file can be a single platform for all project analysis. 29

32 Other Features For those transitioning from SPSS to JMP who are interested in some of JMP s unique features, there are several options available. JMP 9 comes standard with data mining tools such as decision trees and neural networks; these options are not standard in SPSS 19. While data mining tools are advanced features that require some additional reading or training, they can add a completely new level to data analysis. JMP has features for creating reports or dashboards that are very popular in many different industries. As stated earlier, the Graph Builder in JMP 9 contains a mechanism for easy-to-develop graphics that involve map or geographic data. Lastly, JMP has an easy-to-use sample size and power calculator that can be used for simple one and two sample studies involving means and proportions. This platform is easy to use and is useful in research, industry and in classroom discussion on statistical power. 30

33 Conclusion Both SPSS and JMP do a great job of manipulating, visualizing and analyzing data. The purpose of this guide is to transition users who are familiar with SPSS to performing analysis in JMP. We did this through an example that is independent of either software. Looking across several instances, we start to see a pattern emerge in the overall differences JMP and SPSS. Point and click in SPSS is a mechanism for generating SPSS code, so users decide which analysis and options they want to perform and then submit the generated code to obtain the output. The code and results are listed in the output file, which can be saved, copied or manipulated. By contrast, JMP dynamically links the data and reports in order to create an interaction; users start with a general area of analysis and then are allowed to customize output to add different features or analytics. The dynamic link between data and output makes exploring unusual observations very simple, and the interactivity of features such as Graph Builder allows users to create and update visuals in real time. Acknowledgements I would like to thank Curt Hinrichs and Jian Cao at JMP for their help during both the discussion and revision phase of this white paper. I would also like to extend a special thanks to Dr. F. Michael Speed at Texas A&M University, who made some extremely valuable contributions and insight to an early outline of this paper. 31

34 SAS Institute Inc. World Headquarters To contact your local JMP office, please visit: SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. indicates USA registration. Other brand and product names are trademarks of their respective companies. Copyright 2011, SAS Institute Inc. All rights reserved _S