Skip to content

Latest commit

 

History

History
65 lines (49 loc) · 4.04 KB

README.md

File metadata and controls

65 lines (49 loc) · 4.04 KB
Raw

IBM SPSS® Statistics Modeler with Python

  by Jeffrey Chijioke-Uche, MSIS, MSIT, CPDS (IBM Sr. Solution Architect, Hybrid Cloud & Multicloud)
  IBM Information Technology PhD Scholar at Harvard University & Walden University

Usage:

Usage is typically with data analysis for collected non-textual/thematic data. For, example, it could be used when performing dependent sample t-tests where you typically need to determine the following two hypotheses: Null hypothesis (H0), that is if the true mean difference is equal to zero (between the observations) OR Alternative hypothesis(H1), that is if the true mean difference is not equal to zero (two-tailed). With SPSS Modeler statistics flows, you can quickly develop predictive models using business expertise to improve decision making at hypothetical level.

Basics

  1. [CDF ]: CDF plots for a random distribution

More advanced topics

  1. [Confidence intervals]
    1.1 Standard confidence intervals for normal distribution
    1.2 Bootstrapped confidence intervals
    1.3 Bayesian estimates

  2. [Rejection sampling] A method to sample a random distribution

  3. [Binomial distribution] Binomial distribution and Bayesian theorem.

  4. [Power estimation]
    4.1 Standard solver
    4.2 Bootstrapping

Testing hypotheses

Test normality on a distribution

  1. [Normality tests ]
    1.1 Q-Q plots
    1.2 Skew and Kurtosis tests
    1.3 Kormogorov-Smirnov test
    1.4 Shapiro-Wilk test
    1.5 Anderson-Darling test

Goodness of fit

  1. For categorical data
    [chi square test]
  2. For 2 sample distributions
    [Kolmogorov-Smirnov test]

Test difference between means

  1. [Parametric tests & Bootstrapping]
    1.1 t-test
    1.2 Cohen's d (effect size)
    1.3 Bootstrapping

  2. [Non parameteric tests ]
    1.1 Wilcoxon rank-sum test
    1.2 Mann-Whitney test

Test difference between means for dependent groups (repeated measures)

  1. Parametric tests
    1.1 [Paired t-test]
    1.2 [Repeated Measures ANOVA]
  2. Non-parametric tests
    2.1 [Friedman chi square test]

Test percentage change

  1. [Delta method (A/B testing)]