IBM SPSS Statistics Subscription

IBM SPSS Statistics Subscription is the statistical software allowing to solve a set research and business challenges using means of the special analysis, testing of hypotheses, the geospatial and predictive analysis.

IBM SPSS Statistics is applied for analysis data and trends, forecasting and planning, verification of the assumptions and preparation of valid conclusions. The solution can be useful for work of analysts, managers, heads of sales departments IT , etc.

2019

Start of sales on the MerliONCloud platform

On April 18, 2019 the MERLION company, the Russian VAD distributor, within development of the Merlioncloud platform, announces start of sales of two products - IBM SPSS Statistics Subscription and IBM Watson Studio Desktop Subscription. Read more here.

One-component dispersion analysis with repeated measurements

In April, 2019 there was the next IBM SPSS Statistics Subscription software updating.

Procedures of the analysis

• Quantile regression:
  • Models interrelation between a set of variable predictors (independent) and certain percentiles (or "quantiles") a target variable (dependent), most often a median. You watch the additional information in the section Quantile regression.
  • Quantile regression does not make the assumptions of distribution of a target variable, shows a tendency to compensate influence of observations emissions and is widely used for researches in practical areas, such as ecology, health care and financial economy.

• ROC analysis:
  • Estimates the accuracy of predictions of model by construction of the diagram of sensitivity depending on value (1 minus specificity) the test of classification (as the threshold is different on all range of results of diagnostic test). The ROC analysis supports calculation of data of the area under a curve, curves accuracy completeness (precision-recall, PR) and options for comparison of two ROC curves generated or for independent groups, or for pair objects. The additional information you watch the ROC analysis in the section.

• Bayesian statistics:
  • One-component dispersion analysis with repeated measurements is added. This procedure measures one factor of the same object in each separate timepoint or under each condition. It is supposed that for each object there is the only observation for each timepoint or a condition (i.e. interaction of processing of objects is not considered).
  • Improvements of the only binomial selection. In the one-selective procedure of a Bayesian inference (for the binomial distributed data) it is possible to take binomial distribution as a basis. Parameter π, success probability is considered at this number of tests which can terminate in success or failure. Tests do not depend from each other, and probability π remains invariable in all tests. The binomial distributed random variable can be considered the amount of this number of the independent tests submitting to Bernoulli's distribution.
  • Improvements of the only Poisson selection. In the one-selective procedure of a Bayesian inference (for the binomial distributed data) it is possible to take Poisson's distribution as a basis. Poisson's distribution — useful model for rare events; in it it is supposed that for short intervals the probability that the event will come during this interval, is proportional to waiting time. When Bayesian statistical inference is drawn on the basis of Poisson's distribution, the integrated a priori distribution is selected from family of Gamma distributions.

• Reliability analysis:
  • The procedure is updated and now provides options for statistics of a kappa with several certifying for Fleys which estimate the consent of mezhrespondentny estimates to define reliability among different experts. Higher consent provides higher rate of trust in the ratings reflecting true circumstances. Options with several certifying for Fleys are available to statistics of a kappa in the Reliability analysis dialog box: Statistics.

Improvements of commands

• GENLINMIXED command:
  • Structures with type of a covariation matrix ARH1 & CSH, Random effects. The options CSH and ARH1 are added to subcommand/RANDOM (a key word of COVARIANCE_TYPE).
  • Structures with type of a covariation matrix of ARH1 & CSH, the Repeating effects. The options CSH and ARH1 are added to subcommand/DATA_STRUCTURE (a key word of COVARIANCE_TYPE).
  • Method of degrees of freedom of Kenward-Roger. The option KENWARD_ROGER is added to subcommand/BUILD_OPTIONS (a key word of DF_METHOD).
  • Types of covariance of Kronecker. The options UN_AR1, UN_CS, UN_UN are added to subcommand/DATA_STRUCTURE (a key word of COVARIANCE_TYPE).
  • Create a key word of KRONECKER_MEASURES. The key word is used for determination of the list of variables for subcommand/DATA_STRUCTURE. The key word should be used, only when COVARIANCE_TYPE is one of three types of Kronecker. Rules for KRONECKER_MEASURES match REPEATED_MEASURES. When both specifications work, at them can be, and there can not be no general fields, but their values cannot match in accuracy (irrespective of coincidence or mismatch of their order).

• MIXED command:
  • In the subcommand of CRITERIA the key word of DFMETHOD is entered.
  • In the subcommand of REPEATED the key word of KRONECKER is added. The key word should be used, only when COVTYPE is one of three following types of Kronecker.
  • The options UN_AR1, UN_CS and UN_UN are added to a key word of COVTYPE of the subcommand of REPEATED.

2017: Support of a Bayesian statistics

As a result of release of the update of August, 2017 SPSS Statistics began to support a Bayesian statistics. A Bayesian inference - a method of statistical inference in which the Bayes' theorem how hypothesis probability after obtaining the additional information changes is used.

The following Bayesian statistics are supported:

• One-selective student criterions and two-sample student criterions for dependent selections
• Binomial proportional criteria with one selection
• Analysis of distribution of the only selection, Poisson
• Connected selections
• t-criterion for independent selections
• Paired correlations (Pearson)
• Linear regression
• One-component dispersion analysis
• Loglinear regression

Optimization of print preview of an output "Copies as"

Now it is possible to right-click on the selected object in print preview of an output and to select Editing> to Copy as to make the copy of this object in certain formats (for example, Everything, the Image or the Graphic object of Microsoft Office). If to select Editing> to Copy, then Everything will be copied.