Copyright © 2014, SAS Institute Inc. All rights reserved.
MASTERING JMP:
Statistical and Predictive Modelling
Dr. Malcolm Moore
JMP Systems Engineering Manager, Europe
Malcolm.Moore@jmp.com
Copyright © 2014, SAS Institute Inc. All rights reserved.
TOPICS COVERED
Copyright © 2014, SAS Institute Inc. All rights reserved.
WHAT IS A STATISTICAL MODEL?
 An empirical model that relates a set of inputs (predictors,
Xs) to one or more outcomes (responses, Ys).
 Separates the response variation into signal and noise:
Y = f(X) + E
• Y is one or more continuous or categorical response
outcomes.
• X is one or more continuous or categorical predictors.
• f(X) describes predictable variation in Y (signal).
• E describes non-predictable variation in Y (noise).
Copyright © 2014, SAS Institute Inc. All rights reserved.
IDENTIFYING A USEFUL
STATISTICAL MODEL
 “All models are wrong, but some are useful”,
George Box
 Practically what does this mean?
• Guard against over-fitting or assigning too much of the variation
in Y to f(X)?
• Likely to obtain data in future that disagrees with model  revise
model
Copyright © 2014, SAS Institute Inc. All rights reserved.
LEARNING IN FACE OF UNCERTAINTY
IS OFTEN INCREMENTAL
5
What we
think is
happening
Measurement
and Data
Collection
Situation
Appraisal
Measurement
and Data
Collection
What is
really
happening
Able to Consistently Meet Customer Requirements
ModelReal World
Unable to Consistently Meet Customer Requirements
Y = F(X) + Error
Measurement
and Data
Collection
Situation
Appraisal
Situation
Appraisal
Copyright © 2014, SAS Institute Inc. All rights reserved.
MODEL BUILDING IN JMP®
Copyright © 2014, SAS Institute Inc. All rights reserved.
MODEL BUILDING TIPS
 Effective model building
• EDA
• Statistical Methods
• Knowledge/Experience
− Can take some time to master
 A simpler and robust workflow would help more
users solve more problems faster and easier
Copyright © 2014, SAS Institute Inc. All rights reserved.
SCALABILITY?
 What about larger data or complex problems?
• Many Xs
• Correlated Xs
• Outliers or Wrong Values
• Missing Cells
 Need for a scalable, simple and robust workflow for
statistical modelling
• Help a larger community of users get value from statistical
modelling quickly and simply
Copyright © 2014, SAS Institute Inc. All rights reserved.
MODEL BUILDING IN JMP® PRO
Copyright © 2014, SAS Institute Inc. All rights reserved.
SUMMARY
 JMP highly effective statistical modelling tool
• EDA
• Statistical Methods
• Knowledge/Experience
 JMP Pro simplifies statistical modelling
• Scales to any number of Xs and Ys
• Effective with messy data
• Makes statistical modelling simpler and faster
• Reduces risk of selecting “wrong” model
− Make correct decisions faster
− Avoid solving same problem again and again …

Statistical and Predictive Modelling

  • 1.
    Copyright © 2014,SAS Institute Inc. All rights reserved. MASTERING JMP: Statistical and Predictive Modelling Dr. Malcolm Moore JMP Systems Engineering Manager, Europe Malcolm.Moore@jmp.com
  • 2.
    Copyright © 2014,SAS Institute Inc. All rights reserved. TOPICS COVERED
  • 3.
    Copyright © 2014,SAS Institute Inc. All rights reserved. WHAT IS A STATISTICAL MODEL?  An empirical model that relates a set of inputs (predictors, Xs) to one or more outcomes (responses, Ys).  Separates the response variation into signal and noise: Y = f(X) + E • Y is one or more continuous or categorical response outcomes. • X is one or more continuous or categorical predictors. • f(X) describes predictable variation in Y (signal). • E describes non-predictable variation in Y (noise).
  • 4.
    Copyright © 2014,SAS Institute Inc. All rights reserved. IDENTIFYING A USEFUL STATISTICAL MODEL  “All models are wrong, but some are useful”, George Box  Practically what does this mean? • Guard against over-fitting or assigning too much of the variation in Y to f(X)? • Likely to obtain data in future that disagrees with model  revise model
  • 5.
    Copyright © 2014,SAS Institute Inc. All rights reserved. LEARNING IN FACE OF UNCERTAINTY IS OFTEN INCREMENTAL 5 What we think is happening Measurement and Data Collection Situation Appraisal Measurement and Data Collection What is really happening Able to Consistently Meet Customer Requirements ModelReal World Unable to Consistently Meet Customer Requirements Y = F(X) + Error Measurement and Data Collection Situation Appraisal Situation Appraisal
  • 6.
    Copyright © 2014,SAS Institute Inc. All rights reserved. MODEL BUILDING IN JMP®
  • 7.
    Copyright © 2014,SAS Institute Inc. All rights reserved. MODEL BUILDING TIPS  Effective model building • EDA • Statistical Methods • Knowledge/Experience − Can take some time to master  A simpler and robust workflow would help more users solve more problems faster and easier
  • 8.
    Copyright © 2014,SAS Institute Inc. All rights reserved. SCALABILITY?  What about larger data or complex problems? • Many Xs • Correlated Xs • Outliers or Wrong Values • Missing Cells  Need for a scalable, simple and robust workflow for statistical modelling • Help a larger community of users get value from statistical modelling quickly and simply
  • 9.
    Copyright © 2014,SAS Institute Inc. All rights reserved. MODEL BUILDING IN JMP® PRO
  • 10.
    Copyright © 2014,SAS Institute Inc. All rights reserved. SUMMARY  JMP highly effective statistical modelling tool • EDA • Statistical Methods • Knowledge/Experience  JMP Pro simplifies statistical modelling • Scales to any number of Xs and Ys • Effective with messy data • Makes statistical modelling simpler and faster • Reduces risk of selecting “wrong” model − Make correct decisions faster − Avoid solving same problem again and again …