# Analyzing Categorical Data in JMP Melinda Thielbar Senior

Analyzing Categorical Data in JMP Melinda Thielbar Senior Research Statistician Developer, JMP Division, SAS EXAMPLE DATA A traditional retail business has just started an online store. Counts and Multiple Responses They have two goals: 1) Reach new customers who may not be close to their physical locations. 2) Offer more variety in the online space than they can in a physical space. urvey Questions: Ordinal/Likert Scale Com pan y Con fide n t ia l For In t e rn a l Us e On ly

Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed. Exploring Categorical Data Bar charts convey a lot of information, and they are easy to read One goal for the online store was to provide more Selection. Online shoppers dont seem any happier with the Selection than instore shoppers. In fact, none of our online shoppers rated Selection as a 4 or 5, and very few of the people who use both rated Selection as 5. Com pan y Con fide n t ia l For In t e rn a l Us e On ly Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed. Exploring Categorical Data Mosaic plots let you see the relationship between two categorical variables

Height is determine d by the count of the Y-axis variable. Both are ordinal 1-5 variables. The stair steps show that they are related. Customers seem to have a positive or negative overall impression. If they are unhappy with Service, they are also unhappy with Selection. Width is determined by the count for the X-axis variable. Com pan y Con fide n t ia l For In t e rn a l Us e On ly

Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed. Exploring Relationships The Categorical platform produces Likert and share charts Overall, customers are slightly less happy with Selection than they are with Service. The Likert chart gives another view of the same relationship as the mosaic plot. Com pan y Con fide n t ia l For In t e rn a l Us e On ly Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed. Exploring Categorical Data Treemaps show part-to-whole relationships between continuous and

categorical data Largest block implies the largest share of Total Spend. This treemap shows that most of our total sales in dollars come from people who shop both online and in store. Online sales are the smallest share of total sales. Com pan y Con fide n t ia l For In t e rn a l Us e On ly Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed. Modelling Categorical Outcomes Yes/No, Likert Scales, and Counts have their own models Type of Data Which Model

Yes/No Nominal Logistic Likert Scale Ordinal Logistic Count Data Poisson (JMP Pro) Count Data with 0 ZIP Poisson (JMP Pro) Counts Com pan y Con fide n t ia l For In t e rn a l Us e On ly Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed. Modelling Yes/No and Ordinal Data Just a little math

A Nominal Logistic Regression models the probability of a Yes as a function of the input variables. But the data only tells us if each person was a yes or a no. JMP infers the best fit for the relationship between the Xs and the Y. Com pan y Con fide n t ia l For In t e rn a l Us e On ly Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed. Set Up Your Data So JMP Can Help You JMP reads the modelling type of your data and suggests models This is automatically populated when you select a Nominal variable as the Y. If you have an Ordinal

outcome (Likert Scale), JMP automatically selects the Ordinal personality. Setting your modelling types allows JMP to guide you in selecting the correct m Com pan y Con fide n t ia l For In t e rn a l Us e On ly Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed. JMP Reports Are Consistent Across Models People who are farther away shop instore less often. Main driver of satisfaction with selection: Satisfaction Main drivers of online vs offline shopping: Distance to location Com pan y Con fide n t ia l For In t e rn a l Us e On ly

Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed. Summary FIT MODEL HAS PERSONALITIES THAT ARE TAILORED TO THE TYPE OF DATA EXPLORATION WITH GRAPH BUILDER JMP has specialised models for categorical Y variables. Bar charts, mosaic plots, and treemaps are specialized for categorical data. Treemaps show part to relationships. SET YOUR MODELINGwhole TYPES: HELP

JMP HELP YOU JMP JMP Pro Ordinal Nominal Continuous Com pan y Con fide n t ia l For In t e rn a l Us e On ly Co pyr igh t S AS In s t it u te In c. All r igh ts re s er v ed.

## Recently Viewed Presentations

• Counting to ten or other forms of mental games give us a chance to slow our bodies and minds and regain focus so we can regain control over the task at hand. There are many types of mental games, talk...
• To achieve a pass grade: Candidates will produce a portfolio of evidence that includes a proposal scenario, a product that is between 1 and 2 minutes long and includes an interactive menu and splash screen.
• Statistical machine learning. Gather much training data. Design a statistical model. Train, train, train. Given new instances, use model to categorize or predict
• İslami Finansın Birleşik Krallık'ta Gelişimi. Professor Humayon Dar. Yönetim Kurulu Başkan, Başkan & CEO. Edbiz Consulting Ltd. 'Edbiz Danışmanlık Ltd.) [email protected]
• Durable Income is a broad category and may often overlap with fixed income - when fixed income investing is done properly. Durable Income preserves the attractive features of the fixed income asset class (stable, consistent, predictable returns) but potentially mitigates...
• The Tile-HCAL, studies and time plans, V. Korbel, DESY, 30m The last meetings since ECFA-DESY workshop in Prague (Nov. 2002) actual status of R&D: scintillators WLS-fibres, clear fibres, fibre coupling optimisation of tile-fibre system (TFS) photodetectors preamps measurements with the...
• For example, the malate can be transported into the mitochondria via the malate shuttle, reenter the tricarboxylic acid cycle, and get oxidized to oxaloacetate (OAA), which can be used for gluconeogenesis . Alternatively, the OAA can be converted to aspartate...
• SSE remote heating and home control through the mobile app - including heating & smart plugs* Personalised data on household energy use / management . Extended to smart plugs & other flexible domestic load . Greater insight into customer energy...