This page was adapted from a web page at the SPSS web page.
We thank SPSS for their permission to adapt and distribute this page via our web site.
SPSS is committed to providing you with the most powerful and up-to-date statistical
procedures because statistics are the core of your analyses. New statistics and
enhancements to the existing statistical procedures are made in every new release of SPSS
software. And, SPSS releases a new version every 12-18 months. This white paper lists the
statistical enhancements that we’ve added to the SPSS product line starting with version
5.0 through our upcoming version, 9.0. It demonstrates our commitment to the "Real
Stats" part of "Real Stats. Real Easy."
SPSS 5.0
New features
- SPSS CHAID as an add-on module
- PRELIS/LISREL 7 as an add-on module
- Cox Regression procedure which performs proportional hazard regression with time to
response as the dependent variable - Kaplan-Meier procedure which estimates the length of time to an event
- Comprehensive set of integrated statistical charts plus standard business charts
- Levene Test tests of homogeneity in ONEWAY
Enhancements
- MANOVA:
- Ability to request observed power values based upon fixed-effect assumptions for all
univariate and multivariate F tests and t-tests Four types of power values: 1. Approximate
power values, 2. exact power values, 3. alpha level at which the power is to be calculated
for F tests, 4. alpha level at which the power is to be calculated for t-tests Obtain
simultaneous confidence intervals for each parameter estimate and regression coefficient.
Both univariate (Scheffe and Bonferroni) were made available. Request either joint or
individual univariate and multivariate confidence intervals and also vary the confidence
level. Display the effect size values and optimal Scheffe contrast coefficients for the
estimated size of the effects in the model Hierarchical and K-Means clustering methods in
QUICK CLUSTER. Changed the weighted regression method to a nonlinear numerical method to
obtain the correct maximum likelihood estimates in PROBIT. Added discriminant score to the
Discriminant Analysis procedure. Changed the default to two-tailed significance in
CORRELATIONS.
SPSS 6.0
- New features
- Fifty-eight new functions added to the transformation language, including distribution
functions, inverse distribution functions and random number generation functions - Doubled the number of charts available to include Pareto, control charts, error bar,
difference line, high/low/close, range bar, time series/sequence plots, and curve fitting - Third-party API for integration of 3rd party applications/macros into SPSS and its menus
- Fifty-eight new functions added to the transformation language, including distribution
- Enhancements
- Missing value imputation in line charts
- User-defined confidence intervals in EXAMINE
- Display coefficients in user-defined order in regression
- Display multiple reference lines on all charts
- Derived (2nd) scale axis to all bar/line/area charts
SPSS 6.1
- New features
- GLM-based loglinear for fitting loglinear and logit models to count data. GENLOG handles
a new model assumption, POISSON, as well as the current MULTINOMIAL. In addition, GENLOG
can handle messy data situations and accommodates structural zeros. High-resolution
diagnostic plots are integrated into the procedure. - 32 bit back-end for faster statistical processing times
- SPSS Exact Tests as an add-on module integrated with SPSS crosstabs and non-parametric
tests (contact SPSS for available white paper) - One sample t-test
- Grouped median in GRAPH
- Case weights in NPPLOT
- GLM-based loglinear for fitting loglinear and logit models to count data. GENLOG handles
- Enhancements
- Display censored cases in Kaplan-Meier
- Additional asymptotic tests in Chi-square
SPSS 7.0
- New features
- New general linear model procedure (GLM) allows mixed model ANOVA among others. It
contains post-hoc tests and four types of sums of squares. (contact SPSS for available
white paper) - Multidimensional pivot tables for more flexible analysis (white paper available)
- New distribution functions of Lognormal, Logistic, Exponential, Weibull, Gamma, Beta,
Uniform, Pareto, Laplace, and Half Normal in NPPLOT
- New general linear model procedure (GLM) allows mixed model ANOVA among others. It
- Enhancements
- Post-hoc tests in ONEWAY
- Confidence intervals for exponential function parameters in Logistic Regression
- Display the number of cases in an analysis and variable labels in Factor analysis
SPSS 7.5*
- New features
- Amos for structural equation modeling available as an option (white paper available)
- Variance Component Estimation procedure which provides a wide range of estimation
methods to estimate the variance component for each random effect in a mixed model. It
includes four estimation methods: ANOVA, MINQUE, Maximum Likelihood, Restricted Maximum
Likelihood. Includes Type 1 and Type 3 Sums of Squares for ANOVA method. Choices of zero
weight or uniform weight methods. Choice of Maximum Likelihood and Restricted Maximum
Likelihood calculation methods including Fisher’s scoring method and Newton-Raphson. The
ability to save variance components estimates and covariance matrix. (white paper
available) - Missing values analysis is planned as an option. It provides descriptive information for
each unique missing pattern and displays the location of missing values. It also provides
the EM and regression methods to estimate the descriptive statistics of mean, covariance
and correlation among variables of interest when missing data exists.
- Enhancements
- General linear model (GLM) procedure includes pairwise comparisons of expected marginal
means, user-specified error term in post-hoc analysis, linear hypothesis testing of an
effect vs. a linear combination of effects, the option to save design matrix and effect
file, allow unlimited number of factors and fractional numbers in Lmatrix, Mmatrix and
Kmatrix subcommand. (white paper available) - Factor analysis includes covariance matrix for three extraction methods (Principal
Component, Principal Axis, and Image), Promax rotation, Post-rotational sum-of-squares,
ability to apply solution to new cases - Regression creates a system file containing parameter estimates and their covariance and
correlation matrix which can be used for further analysis and prediction - Logistic includes Pseudo-R Squared measures of Cox and Snell’s and Nagelkerke’s. User
specified cutoff value for classification table. The default contrast changed to Indicator
from Deviation, and the addition of Hosmer-Lemeshow goodness-of-fit statistic. - McNemar’s test in crosstabs
- Cross validation estimation of the misclassification rate added to Discriminant
- Exponential of Beta in GENLOG
- Pie chart in Frequencies procedure
- Cluster bar chart in Crosstab procedure
- General linear model (GLM) procedure includes pairwise comparisons of expected marginal
SPSS 8.0*
- Easily determine if differences between multiple groups are statistically significant in
your experiments with a dramatically improved ANOVA. For example, you can test different
price points in a direct mailing. Use as many variables as you like as you create custom
models without limits on maximum order of interaction, and work faster now because you
don’t have to specify ranges of factor levels- SPSS uses the default levels. As you work,
you can pinpoint effects as you perform post-hoc analyses (in ONEWAY) that allow you to
continue your discovery. Then, choose the model that’s right for you as you analyze the
model with four types of sum of squares, and increase your certainty with better data
handling in empty cells, and perform lack of fit tests to select your best model. - Easily compare variance between groups with Robust Levene’s test in EXAMINE. For
example, compare groups with very different variance such as cancer rates in smokers and
non-smokers. - Make better models regardless of your sample sizes with Harmonic and geometric means in
MEANS. For example you can use Harmonic means when your data set has more of one group
than another, or you have two samples of different sizes. - More accurately work with exponential distribution with the Kolmogorov-Smirnov test, as
found in financial or bacterial growth data. - More freely analyze data with an updated interface for commands in GLM that frees you
from using syntax. - Get better models from SPSS Professional Statistics and SPSS Advanced Statistics. In
SPSS Professional Statistics, you can use Intraclass correlation in RELIABILITY. In SPSS
Advanced Statistics, now you can choose the model you like best, by including plots of the
one minus cumulative survival function in SURVIVAL, Kaplan-Meier and COX REGRESSION. And
you can select univariate and multivariate lack of fit tests in GLM to help you test
nested models. - Get more useful information out of your data with SPSS Categories. SPSS Categories is
now two modules, SPSS Conjoint and SPSS Categories, at no extra cost. SPSS Conjoint
includes orthoplan, plancards, and conjoint procedures. The SPSS Categories option offers
two new procedures which free you from data constraints such as yes/no answers and allow
you to analyze any categorical data. Explore the relationship between two categorical
variables with Bi-plots. Create new, modern output that’s ready to send to clients. The
new SPSS Categories includes:- CATREG; perform multiple regression analysis for categorical data
- CORRESPONDENCE; which improves upon ANACOR
SPSS 9.0
- New features
- Predict categorical outcomes with more than two categories with the Multinomial Logistic
Regression in SPSS Regression Models (formerly named SPSS Professional Statistics). For
example, model what predicts whether the customer buys product A, product B or product C. - Make more informed decisions by evaluating the diagnostic performance of a test using
ROC (Receiver-Operating Characteristic) analysis
- Predict categorical outcomes with more than two categories with the Multinomial Logistic
- Enhancements
- A new post-hoc test for repeated measures in General Linear Models (GLM) in SPSS
Advanced Models (formerly SPSS Advanced Statistics) that allows you to compare the
between-groups effect averaged across all time points. For example, how do racial/ethnic
groups vary across all time points combined. - Go beyond the limits of a two-way crosstab as you explore three-way relationships in
categorical data with Cochran-Mantel-Haenszel statistic in Crosstabs.
- A new post-hoc test for repeated measures in General Linear Models (GLM) in SPSS
*Features are subject to change as development progresses
This page was adapted from a web page at the SPSS web page.
We thank SPSS for their permission to adapt and distribute this page via our web site.