Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
45 views

Logistic Regression

Logistic regression was used to predict successful enrolment into a Masters of Psychology program based on interest in the program, average previous degree score, and whether the applicant held a psychology degree. The logistic regression model with all three predictors was able to correctly predict enrolment 66.7% of the time, a 16.7% improvement over the baseline model without predictors. Only previous degree score was a significant predictor, with every unit increase associated with a 17.4% higher chance of enrolling in the program. Assumption testing found no issues with multicollinearity, linearity of logit, independence of errors, or outliers.

Uploaded by

Safaa Kahil
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
45 views

Logistic Regression

Logistic regression was used to predict successful enrolment into a Masters of Psychology program based on interest in the program, average previous degree score, and whether the applicant held a psychology degree. The logistic regression model with all three predictors was able to correctly predict enrolment 66.7% of the time, a 16.7% improvement over the baseline model without predictors. Only previous degree score was a significant predictor, with every unit increase associated with a 17.4% higher chance of enrolling in the program. Assumption testing found no issues with multicollinearity, linearity of logit, independence of errors, or outliers.

Uploaded by

Safaa Kahil
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 30

Logistic Regression

Learning Centre
Content

1 2 3
What is logistic Types of Worked
regression? logistic example on
regression binary logistic
regression
What is Logistic Regression?

Like the multiple regression, logistic regression is a statistical analysis


used to examine relationships between independent variables
(predictors) and a dependant variable (criterion)

The main difference is in logistic regression, the criterion is nominal


(predicting group membership). For example, do age and gender predict
whether one signs up for swimming lessons (yes/no)
Types of Logistic Regression

• There are primarily 2 types of logistic regression: (1) Binary and (2)
Multinomial models. The difference lies in the types of the criterion
variable
• Binary logistic regression is for a dichotomous criterion (i.e., 2-level
variable)
• Multinomial logistic regression is for a multicategorical criterion
(i.e., a variable with more than 2 levels)
• This set of slides focuses on binary logistic regression
Example…
A researcher would like to find out if the three predictors can predict
successful enrolment into the Masters of Psychology program at JCU. The
researcher recruited 30 participants who applied for the program, and asked
them the following questions:

1. Interest in the Masters of Psychology program (rated 1-100)


2. Average overall score from a previous degree (scored 1-100)
3. Holding a psychology degree (yes/no)
4. Successful enrolment (yes/no) Note that sample size of
30 was used only for
A binary logistic regression was then conducted. illustration purposes, an
actual study would require
larger sample size!
Location of SPSS Data Files for Practice

Example SPSS data for practice are available on LearnJCU:

Log in to LearnJCU -> Organisations -> Learning Centre JCU Singapore ->
Learning Centre -> Statistics and Maths -> SPSS Data for Practice
Assumptions Testing

01 03
Multicollinearity Logit Linearity

02 04
Independence of
Outliers
errors
Assumptions Testing
Please refer to the SPSS guide on Multiple Regression of how to conduct
the four assumption tests at
https://www.jcu.edu.sg/current-students/student-support-services/learni
ng-support/statistics-and-mathematics-support
Assumptions Testing

01 03
Multicollinearity Logit Linearity

02 04
Independence of
Outliers
errors
Assumptions Testing: Logit Linearity

This is an assumption that the relationship between


each continuous predictor and a criterion is linear.
● Interest and PreviousScore are continuous, thus
they have to be tested for this assumption.
● PsychDegree is categorical, hence it is not requited
to be tested.
To test for this, we first need to create new variables
in our dataset: Logit functions of the continuous IVs
● Transform  Compute Variable
Assumptions Testing: Logit Linearity

1. Select ‘Arithmetic’ under Function group,


and double click on ‘Ln’ under Functions
and special variables
2. LN should appear under Numeric
Expression
3. To create the logit expression of the first
continuous variable (Interest), double
click on Interest
4. Name the target variable – LnInterest
5. OK
Assumptions Testing: Logit Linearity

Repeat the procedures, this


time creating the logit
function of the other
continuous variable
(PreviousScore)
Assumptions Testing: Logit Linearity

You will see these 2 new


variables in your dataset
Assumptions Testing: Logit Linearity

To conduct the assumption test


for logit linearity, go to Analyze -
> Regression -> Binary Logistic
Assumptions Testing: Logit Linearity

1. Move ‘ Enrolled’ into the Dependent


box
2. Move ‘Interest’ and ‘PreviousScore’ into
the Covariates box
3. Holding the Ctrl key, then select
‘Interest’ and ‘LnInterest’, and click on
‘>a*b>’ to enter the interaction term into
the Covariates box
4. Repeat Step 3 for ‘PreviousScore’ and
‘LnPreviousScore’
Assumptions Testing: Logit Linearity

● You should have 4


Covariates in total

● Click OK
Assumptions Testing: Logit Linearity

Since the p values of the interaction terms are above .05, we conclude
that the assumption for logit linearity is not violated
Logistic Regression

Now to conduct the main


analysis…

Analyze -> Regression ->


Binary Logistic
Logistic Regression

1. Move ‘Enrolled’ into the


Dependent box
2. Move ‘Interest’, ‘PreviousScore’
and ‘PsychDegree’ into the
Covariates box
Note that ‘PsychDegree’ is a categorical
variable.
Logistic Regression

3. Click on Categorical
4. Select ‘PsychDegree’ as a
categorical covariate
5. Continue
Logistic Regression

6. Click on Save
7. Select Probabilities, Group
membership, Cook’s (this can be
used to screen for outliers), and
Standardized Residuals
8. Continue
Logistic Regression

9. Click on Options
10. Select Classification plots,
Hosmer-Lemeshow goodness-
of-fit, and CI for exp(B)
11. Continue, and OK
Assumptions Testing: Outliers

01 03
Multicollinearity Logit Linearity

02 04
Independence of
Outliers
errors
Assumptions Testing: Outliers

● Outliers can be tested


together with the main
analysis
● Looking at the dataset,
Cook’s distance is added
as a new variable
● Since all the values are <
1, we conclude that there
are no outliers
Logistic Regression: Results

● The purpose of logistic regression is thus to find out if the prediction accuracy of
the model can be improved by predictor variables
● This table shows the regression model with no predictors involved (block 0). This
model (at Step 0) can correctly predict if someone successfully enrolled 50%
of the time.
Logistic Regression: Results

In block 1, all the


predictors were entered
simultaneously

R square values of the


regression model
A p value <.05 suggests
that, overall, the predictors
significantly improved the
Measure of model fit. A p
prediction accuracy of the
value >.05 suggests a
model
good model fit
Logistic Regression: Results

In Step 1, the addition of the predictors resulted in the model being able
to predict successful enrolment 66.7% of the time (compared to 50% in
block 0; 16.7% improvement !)
Logistic Regression: Results

● This table tells us which predictors are significant. Only Previous score is a significant predictor (p
< .05)
● In logistic regression, Exp(B) is commonly used to interpret results, and is expressed as an odds
ratio
● In other words, an increase of 1 unit in Previous score results in a 17.4% more chance of enrolling
in the masters program (1.174 – 1 = .174, meaning .174 above 1)
● The other statistics (e.g., B, Wald, 95% CI) can also be reported in the writeup
Write-Up

An example write-up can be found on page 228 in

Allen, P., Bennett, K., & Heritage, B. (2019). SPSS Statistics:


A Practical Guide (4th ed.). Cengage Learning.
Questions?
learningcentre-singapore@jcu.edu.au
 

You might also like