Getting started with regression techniques in SPSS Jarlath Quinn - - PowerPoint PPT Presentation

A SELECT INTERNATIONAL COMPANY

www.sv-europe.com

Getting started with regression techniques in SPSS

Jarlath Quinn – Analytics Consultant Rachel Clinton – Business Development

A SELECT INTERNATIONAL COMPANY

• Premium, accredited partner to IBM specialising in the SPSS Advanced

Analytics suite.

• Team each has 15 to 20 years of experience working in the predictive

analytic space - specifically as senior members of the heritage SPSS team

A SELECT INTERNATIONAL COMPANY

• Overview of regression techniques and linear relationships
• Performing a simple linear regression
• Using multiple linear regression to make predictions
• Predicting response likelihood with logistic regression

Agenda

A SELECT INTERNATIONAL COMPANY

• A family of statistical techniques used to predict outcomes and generate

estimates for hundreds of applications

• Linear regression is used

– When the outcome is continuous (or scale) data – The relationships between the fields can be described using straight lines

• Non linear regression

– Is used when the outcome itself is not linear (such as a category) – Often takes the form of logistic regression

What do we mean by ‘Regression’?

A SELECT INTERNATIONAL COMPANY

Where are regression techniques used?

• Modelling the relationship between promotion spend and revenue
• Estimating pollution levels following heavy rainfall
• Predicting tourism revenue based on exchange rates and air travel costs
• Predicting student test scores based on previous test results and peer-group

performance

• Estimating website hits based on re-tweets and follower numbers
• Predicting sales of barbeques based on temperature forecasts

A SELECT INTERNATIONAL COMPANY

Regression to the mean

A SELECT INTERNATIONAL COMPANY

Regression to the mean

*jittered points

A SELECT INTERNATIONAL COMPANY

Regression to the mean

A SELECT INTERNATIONAL COMPANY

Regression to the mean

A SELECT INTERNATIONAL COMPANY

Regression to the mean

}

The regression line is drawn so it that minimises the differences between the points and line itself

( this is called the line of least squares)

A SELECT INTERNATIONAL COMPANY

Regression to the mean

* *Anscombe’s Quartet

• But be careful…
• It is just an average

after all…

A SELECT INTERNATIONAL COMPANY

Measuring linear relationships

0.859 0.434

• .701

Pearson correlation values

A SELECT INTERNATIONAL COMPANY

Non-linear relationships

• 0.005
• .671

Pearson correlation values

A SELECT INTERNATIONAL COMPANY

Correlations as percentages

• Correlation = 0.859
• 0.859 x 0.859 = 0.738
• 0.738 = 73.8 %
• Correlation Squared = ‘R Square’

73.8%

A SELECT INTERNATIONAL COMPANY

From correlation to prediction

How can we express linear relationships as predictive models?

A SELECT INTERNATIONAL COMPANY

www.sv-europe.com

How long does it take to cook a chicken?

A SELECT INTERNATIONAL COMPANY

How long does it take to cook a chicken?

• 7 minutes per pound plus 45 minutes
• 20 minutes per pound plus 20 minutes

y = mx + c

• r

y = a + bx

A SELECT INTERNATIONAL COMPANY

A SELECT INTERNATIONAL COMPANY

www.sv-europe.com

Lets look at a demo of linear regression in IBM SPSS Statistics

A SELECT INTERNATIONAL COMPANY

www.sv-europe.com

How can we predict category outcomes?

A SELECT INTERNATIONAL COMPANY

Logistic regression

• Allows us to predict things that linear regression can’t
• Such as…

– Response to a marketing campaign – Credit risk – Whether a subscriber is likely to renew a service – Risk of equipment failure – How likely is it that a particular patient will be readmitted to hospital – Whether a charity donor will switch to direct debit

A SELECT INTERNATIONAL COMPANY

Logistic regression

• But….
• These outcomes are not continuous numbers so standard linear regression

won’t work

• When the outcome consists of two categories we use binary logistic

regression

• When the outcome has three or more categories we use multinomial

logistic regression

• Logistic gets around the limitations of describing relationships with straight

lines by using a special sigmoid curve

A SELECT INTERNATIONAL COMPANY

Logistic regression

00

Discount %

Probability of Responding

A SELECT INTERNATIONAL COMPANY

IBM SPSS Regression Module and R integration

• Using the SPSS Regression Module we can go

beyond Linear Regression and unlock many

• ther types of Regression functionality

A SELECT INTERNATIONAL COMPANY

Exclusive Smart Vision bundle offers

Regression add-on pack

• If you already have SPSS base and

add the regression module to your licence… – 1 user perpetual licence + 1st year support + 1 day personalised training – £2,900 (+ VAT) – Saving £500 on IBM list price – Offer code SVRegSP001

SPSS & Regression starter pack

• If you do not have SPSS base licence
• already. Purchase Base & Regression

– 1 user perpetual licence + 1st year support + 2 days personalised training – £4,229 (+ VAT) – Saving £1,000 on IBM list price – Offer code SVRegSP002

A SELECT INTERNATIONAL COMPANY

Working with Smart Vision Europe Ltd

• As a premier partner we sell the IBM SPSS suite of software to you directly

– We’re agile, responsive and generally easier to deal with

• As experts in SPSS / analytics / predictive analytics we will

– Deliver classroom training courses – Offer side by side training support – Offer “skills transfer” consulting – Run booster and refresher sessions to get more from your SPSS licences – Give no strings attached advice

• We are a support providing partner so if you already have SPSS you can source

your technical support directly from us (identical costs to IBM)

– We offer telephone support with real people as well as web tickets / email queries – We offer “how to” support to help you get moving on your project quickly

A SELECT INTERNATIONAL COMPANY

www.sv-europe.com

Thank you

Contact us: +44 (0)207 786 3568 info@sv-europe.com Twitter: @sveurope Follow us on Linked In Sign up for our Newsletter