New Generations of Statistical Applications

As the clock clicks closer to the next millennium, we are reminded that we have to face a new generation of computers and software applications. Yes, they are here! This article continues with a series begun last month in which we introduce the new wave of "next generation" statistical packages. We continue in this issue with SPSS 9.0. Let's first examine what we have received lately:

Most Current Versions of Statistical Applications at RSS

Stat. Application	Windows 95/NT	Mac	UNIX
SPSS	9.0 *	6.1	6.13 #
SAS	7.0 * #	6.12 TS040	7.0 #
S-Plus	4.5	-	5.0 #

red font - newly arrived
* under evaluation
# to be installed

SPSS 9.0

An overall evaluation of any comprehensive computer package is difficult to approach. It seems like I am always learning new things, even about packages that I have used extensively for long periods of time. However, I will do my best in providing a brief overview, while attending to new features.

This review assumes that readers have some experience with SPSS for Windows. If you are familiar with SPSS 7.5 and 8.0, great! Not much has changed. In fact, in my opinion, SPSS has done few major changes since they stopped using ASCII output and went to using an object-oriented approach. Another brief, general comment is that SPSS appears to have accepted a model-based approach to statistical analysis (e.g., focus on general and generalized linear models). For example, the documentation provided by SPSS includes a manual on regression models and one on advanced models. This appears to reflect the approach of the field of data analysis in general, focusing on applications of models as opposed to focusing on specific statistical procedures.

My final general impression is that SPSS continues to move in a direction to support more business-oriented applications as opposed to developing more statistical routines. That being said, I find SPSS a very useful product, and I find some of the new developments timely and helpful.

New Directions

As mentioned above, SPSS 9.0 appears to be moving toward business applications. It appears to be particularly committed to Report On Line Analytical Processing (OLAP) and the SmartViewer Web Server. The impetus behind these developments is to provide a medium in which organizations can create and distribute information accessible by a Web browser that is searchable and editable. OLAP cubes contain information in SPSS table format, but the user can control how that information is displayed.

For example, in examining the relationship between happiness and region of the United States, SPSS creates the table below by default:

OLAP Cubes

Region of the United States: Total

	Sum	N	Mean	Std. Deviation	% of Total Sum	% of Total N
General Happiness	2706	1504	1.80	.62	100.0%	100.0%

However, information is stored in the cube by each region of the United States. To access this information, the researcher can work with the cube in the SPSS Pivot Table window, examining the relationship for each region of the United States (e.g., the West):

OLAP Cubes

Region of the United States: West

	Sum	N	Mean	Std. Deviation	% of Total Sum	% of Total N
General Happiness	749	420	1.78	.61	27.7%	27.9%

As you can plainly see, the California sunshine is vastly overrated.

The Smart Viewer Web Server enables the user to conveniently post this information on the Web, while making tables searchable and editable. These features seem ideal for a business environment, in which users could access and edit information conveniently over the Internet or an intranet; however, the utility appears limited for academic environments. For example, the information that can be presented in an OLAP cube is limited to summary statistics and graphs. Please see the following URLs for more information: http://www.spss.com/reportOLAP, http://www.spss.com/software/spss/sviewer/Web/.

Expanded Graphical Capabilities

SPSS 9.0 expands the graphical capabilities that first became available with the interactive graphics module on version 8.0. Stacked bar charts and area graphs are now available in the interactive graphics module, and multiple response variables can now be graphed in a line graph for instance. SPSS 9.0 Interactive Graphs and Pivot Tables now allow a user to easily convert tables to graphs and vice versa, just right mouse click on the pivot table (see Figure 1).

Figure 1. Converting Pivot Table to Graph in SPSS 9.0.

Data Access

Reading in ASCII or text data becomes much easier with SPSS 9.0. SPSS 9.0 entails the use of a Text Wizard similar to Excel (See Figure 2). This relieves the user from having to read in ASCII data with the rather tedious procedure of specifying variable names and column numbers (in the case of fixed column data entry). In addition, the database capture option in SPSS 9.0 works quite well. With this feature, you can import your Access or Fox Pro databases with little difficulty. Database capture was also available in SPSS 8.0; however, I found it to be very unstable. I do not know if this is something unique to my machine or if the procedure has been improved in SPSS 9.0. SPSS 9.0 also allows the user to directly access powerful databases such as Oracle and Essbase.

Figure 2. Read Text Data Option in SPSS 9.0

New Statistical Procedures

SPSS 9.0 has some new statistical procedures available; however, the development of these applications is limited. The most substantial change is that SPSS now does multinomial logit, or multinomial logistic regression if you prefer. It was somewhat bothersome with SPSS 8.0 to be restricted to a dichotomous outcome variable when doing a logistic regression. This is located under the regression menu (see Figure 3). Other Statistical procedures added to SPSS 9.0 include:

Receiver-Operating Characteristic (ROC) curve analysis

The Cochran-Mantel-Haenszel statistic is available in Crosstabs. This allows a user to test the significance of a 2x2 contingency table, covarying for a third dichotomous categorical variable.

PROXSCAL is an alternative multidimensional scaling program available with Categories 9.0.2. This update is available with SPSS Categories 9.0. We will be making this available when we install SPSS 9.0 on the network servers.

Conjoint analysis is included as a separate module, SPSS 9.0 Conjoint Analysis (see more about this in an upcoming Benchmarks Online Article).

Figure 3. Multinomial Logistic Regression in SPSS 9.0

Conclusion

As mentioned previously, SPSS 9.0 has several new developments. However, my opinion is that most of these developments are more useful for a business context, and will find limited use in an academic setting. My hope is that in future releases, that SPSS will introduce more data analytic methods, such as more graphical capabilities for exploratory data analysis, the ability to do hierarchical linear modeling, and more robust methods. However, in my opinion, SPSS's strength has been and continues to be ease of use. The features that have made it so easy to use, such as the spreadsheet format for SPSS system files, the option to paste syntax to the syntax window, and the intuitive Windows-based point-and-click approach remain with version 9.0, and it is the ease of use that brings me back to using SPSS for most of by data analyses. I hope that you have found this review useful; we will be making SPSS 9.0 available on network servers soon. Following installation on the servers, we will release the CD for personal installation for faculty members. Please contact me at 565-2140 or by E-mail at craigh@unt.edu if I can be of assistance in your data analysis.