## Exclusive to Différence

Designed and programmed by Difference for the specific needs of the industry.

## Change Point Analysis

In today’s environment, we are often buried under data ranging from sales and production figures to absenteeism and accident rates. Fundamental questions we try to answer when we look at a key performance indicator or historical data are: Did changes occur? When? What was their amplitude? Are we improving our performance? Are we sure?

The change point analysis, a versatile tool minimally affected by outliers, can analyze almost any kind of data. It identifies the moment a shift in mean was observed, the magnitude of the shift and the probability the shift is not a ‘real’ shift but just a pattern observed by chance. ## Basics Statistical Analyses

In order to make basic statistical tools accessible and easy to use, we suggest an Excel-based tool showing only the essentials in an easy-to-interpret format. Some of its features are:

• Edit your data using the standard functionalities of Excel
• Get the descriptive statistics of one variable
• Build Individuals Control Chart
• Perform a capability analysis for normal data
• Test one, two means or more means (t-tets, ANOVA)
• Compare paired observations
• Perform linear regression and residual analysis

### Try it!

##### Get a FREE 30-day demo  ##### Analysis with 1 variable
• Descriptive statistics, box plot, histogram, capability analysis, quantiles
• ImR and EWMA control charts with the Western Electric decision rules
• Change Point Analysis (CPA): automatic detection of historical changes in the mean
• Statistical tests on: average, standard deviation, autocorrelation, and confidence intervals ##### Pareto Analysis
• Most frequent classes with cumulative curve - find the 80/20 items!
• Use weights to assess relative importance ##### Distribution Fitting
• For simulations: Exponential, Lognormal, Normal, Triangular, Uniform, Weibull, etc.
• Fit or compare to data using robust procedure and graphical assessments
• Sample from known parameters to get the shape and properties ##### Time Series Forecasting
• Time series diagnosis: stability of the mean, autocorrelation, detection of cycles
• Holt-Winters double exponential seasonal smoothing (also called triple exponential)
• Autoregressive PLS: robust autoregressive and seasonal models ##### Means comparison and ANOVA
• T Test for two independent groups and for paired observations
• F-test for several groups with Boneferroni all pairs comparison
• Power and sample size calculations using non-central T and F distributions ##### Simple Regression (one X)
• Correlation and density ellipse
• Linear regression, with polynomial term and standard transforms, residual analysis
• Nonlinear regression: built-in power law, exponential, growth & logistic S-shape models ##### Contingency Analysis
• Mosaic Plot to analyze crossed frequencies (Marimekko chart)
• 2D Frequency table - find the 80/20 items!
• Use weights to assess relative importance ##### Plan a Measurement System Analysis
• Design the experiment: specify testers and samples
• Prepare the design: add replicates and randomize trials
• Generate data collection sheet with multiple measurements for the same treatment ##### Analyze Measurement System Results
• Variance decomposition: unbalanced random effect ANOVA, handling of negative variances
• Results visualization: multi-vari plots, S control chart, interaction plot
• Measurement usefulness: increment diagnosis, comparison to total or historical variation
• Risk analysis: Monte-Carlo simulation to design acceptance limits and understand decision-making risk ##### Monte-Carlo Simulation
• Decision variables or assumptions: choice of statistical distributions, truncation limits
• Calculated responses: random white noise, specification limits, user-defined formulas
• Random sampling: pure random or Latin hypercube on probabilities
• Statistical reports: histograms, descriptive statistics, capability analysis, variable influence ranking
• No programming required!  