Scribd
Upload
32 views17 pages

Multinominal Logistic Regression - D. Boduszek

A multinomial logistic regression was conducted to analyze predictors of recidivism among prisoners categorized as first, second, or third+ time incarcerated. Higher criminal thinking styles significantly increased the likelihood of second or third+ incarceration compared to first incarceration. Living in urban or rural areas also increased the likelihood of a second incarceration compared to living outside the country. Non-violent offenses decreased the likelihood of third+ incarceration compared to violent offenses. Higher criminal friends and thinking styles, as well as living rurally, increased the likelihood of third+ incarceration.

Uploaded by

Tofik
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
32 views17 pages

Multinominal Logistic Regression - D. Boduszek

A multinomial logistic regression was conducted to analyze predictors of recidivism among prisoners categorized as first, second, or third+ time incarcerated. Higher criminal thinking styles significantly increased the likelihood of second or third+ incarceration compared to first incarceration. Living in urban or rural areas also increased the likelihood of a second incarceration compared to living outside the country. Non-violent offenses decreased the likelihood of third+ incarceration compared to violent offenses. Higher criminal friends and thinking styles, as well as living rurally, increased the likelihood of third+ incarceration.

Uploaded by

Tofik
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 17

Daniel Boduszek

University of Huddersfield
d.boduszek@hud.ac.uk
 Introduction to Multinominal Logistic
Regression

 SPSS procedure of MLR


 Example based on prison data

 Interpretation of SPSS output

 Presenting results from MLR


 Tables
 Presenting results in text
 MLR is an extension of binary logistic regression

 MLR is appropriate when the outcome variable is


categorical with more than two categories and the
predictors are of any type: nominal, ordinal and / or
continuous

 Multinomial logistic regression does not require the use


of a coding strategy (i.e. dummy coding) for including
categorical predictors in the model. Categorical
predictor variables can be included directly as factors in
the multinomial logistic regression dialog menu box.
 Variables in the model
 DV = Recidivism (3 categories)
 1 - first time in prison (reference category)
 2 - second time in prison
 3 - third time (or more) in prison
 IVs Categorical
 Type of criminal offence (0 = non-violent; 1 = violent (ref))
 Location of offences (1 = urban; 2 = rural; 3 = outside the country (ref))
 IVs Continuous
 Criminal Friends Index
 Criminal Thinking Style

 The main interest of current analysis is to


focus on the relationship between criminal
thinking & criminal friends and recidivism
while controlling for type and location of
offences.
 From the menu at the top of the screen click Analyze, then select
Regression, then Multinominal Logistic
 Choose you categorical DV (Recidivism_3cat) and move it into the Dependent box
 Move the categorical IVs (Type of Criminals and Location) into Factor(s) box and
continuous IVs (Criminal Friend Index and Criminal Thinking) into Covariate(s) box.

Click on the Reference


Category... button and
select First Category
 Click on the Statistics
button and select the
following

 Click Continue
 Click on the Model button.

 In the Specify Model


section, click on Main
effects.

 Click on Continue.

 And OK
 The Case Processing Summary
table simply shows how many
cases or observations were in
each category of the outcome
variable (as well as their
percentages) and categorical
predictors.

 It also shows if there was any


missing data.
 The Model Fitting Information table shows various indices for assessing the null model and
the final model which includes all the predictors and the intercept (sometimes called the full
model)

 Both the AIC and the BIC are information theory based model fit statistics. Lower values of
indicate better model fit and both can be below zero (i.e. larger negative values indicate
better fit)

 Here, we see model fit is significantχ² (10) = 89.93, p < .001, which indicates our full model
predicts significantly better, or more accurately, than the null model

 To be clear, you want the p-value to be less than your established cutoff (generally 0.05) to
indicate good fit
 The Goodness-of-Fit table provides further
evidence of good fit for our model. Again,
both the Pearson and Deviance statistics are
chi-square based methods and subject to
inflation with large samples.

 Here, we interpret lack of significance as


indicating good fit. To be clear, you want the
p-value to be greater than your established
cutoff (generally 0.05) to indicate good fit.

 The Pseudo R-Square table displays three


metrics which have been developed to
provide a number familiar to those who have
used traditional, standard multiple
regression. They are treated as measures of
effect size, similar to how R² is treated in
standard multiple regression. However,
these metrics do not represent the amount
of variance in the outcome variable
accounted for by the predictor variables.
Higher values indicate better fit.
 The statistics in the Likelihood Ratio Tests table are the same types as those reported for
the null and full models above in the Model Fitting Information table. Here however, each
element of the model is being compared to the full model in such a way as to allow the
research to determine if each element should be included in the full model. In other words,
does each element (predictor) contributed meaningfully to the full effect.
 Table
 The Parameter Estimates table, shows the logistic coefficient (B) for each
predictor variable for each alternative category of the outcome variable.
Alternative category meaning, not the reference category. The logistic
coefficient is the expected amount of change in the logit for each one unit
change in the predictor. The logit is what is being predicted; it is the odds of
membership in the category of the outcome variable which has been
specified (here the first value: 1 was specified, rather than the alternative
values 2 or 3). The closer a logistic coefficient is to zero, the less influence the
predictor has in predicting the logit. The table also displays the standard
error, Wald statistic, df, Sig. (p-value); as well as the Exp(B) and confidence
interval for the Exp(B). The Wald test (and associated p-value) is used to
evaluate whether or not the logistic coefficient is different than zero. The
Exp(B) is the odds ratio associated with each predictor. We expect predictors
which increase the logit to display Exp(B) greater than 1.0, those predictors
which do not have an effect on the logit will display an Exp(B) of 1.0 and
predictors which decease the logit will have Exp(B) values less than 1.0.
The Classification Table shows how
well our full model correctly classifies
cases. A perfect model would show
only values on the diagonal--correctly
classifying all cases. Adding across
the rows represents the number of
cases in each category in the actual
data and adding down the columns
represents the number of cases in
each category as classified by the full
model. The key piece of information is
the overall percentage in the lower
right corner which shows our model
(with all predictors & the constant) is
52.9% accurate.
 Table 1
2nd Incarceration (n=104) 3rd (or more) Incarceration (n=77)

Variable OR (95% CI) SE OR (95% CI) SE

Criminal Friends 1.02 (1.00/1.05) .01 1.07(1.04/1.10)*** .02

Criminal Thinking 1.07(1.03/1.11)*** .02 1.07(1.02/1.12)** .02

Type of offence

Non-violent .85 (.48/1.51) .29 .27(.14/.53)*** .34

Violent 1 1

Location of offence

Urban 2.87 (1.11/7.39)*** .48 1.91(.63/5.86) .57

Rural 3.59 (1.32/9.73)*** .51 5.27(1.66/16.80)** .59

Outside the country 1 1

Note. Reference group: 1st Incarceration (n=131). OR = Odds Ratio. SE = Standard Error. 95% CI = Confidence
Interval. * p<.05; ** p<.01; *** p<.001
 A Multinomial Logistic Regression was used to analyse predictors for an unordered group
classification, such as prisoners who were incarcerated for the first time, prisoners who were
incarcerated for the second time, and prisoners who were incarcerated for the third time (or
more). The reference category for the outcome variable was ‘first incarcerated prisoners’; each of
the other two categories was compared to this reference group. The main interest of current
analysis was focused on the relationship between criminal thinking & criminal friends and
recidivism (3 categories) while controlling for type and location of offences.
 The first column in Table 1 has the outcome of “second incarceration” compared to “first
incarceration” (reference category). The results suggest that criminal friends have no significant
effect on the recidivism. However, higher levels of criminal thinking style (OR = 1.07) significantly
increase the probability of recidivism. In relation to the location of offences participants from
urban (OR = 2.87) and rural areas (OR = 3.59), compared to offences committed outside of the
country, are more likely to be incarcerated more than once. Type of offence was not a significant
predictor of recidivism.
 The second column in Table 1 has the outcome of “third (or more) incarceration” compared to
“first incarceration” (reference category). Statistical analysis shows that those participants who
reported higher level of criminal thinking (OR = 1.07) and associations with criminal friends (OR =
1.07) were significantly more likely to report recidivism. Non-violent offences (OR = .27),
compared to violent offences decrease the probability of the recidivism. In relation to the
location of offences participants from rural areas (OR = 5.27), compared to offences committed
outside of the country, are over five times more likely to be incarcerated on 3 or more occasions.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy