Scribd
Upload
3 views

Multinomial Logistic Regression

The document discusses multinomial logistic regression, a statistical method used to model nominal outcome variables with more than two categories. It explains the model's application, mathematical formulation, and provides an example involving high school students' program choices based on socioeconomic status and writing scores. The document also highlights the interpretation of coefficients and the relative odds associated with different predictor variables.

Uploaded by

amanvocational
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
3 views

Multinomial Logistic Regression

The document discusses multinomial logistic regression, a statistical method used to model nominal outcome variables with more than two categories. It explains the model's application, mathematical formulation, and provides an example involving high school students' program choices based on socioeconomic status and writing scores. The document also highlights the interpretation of coefficients and the relative odds associated with different predictor variables.

Uploaded by

amanvocational
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 17

 When to select the model

 Reasons for the popularity of the model


 The mathematical form of the model
 Its Many applications
 Multinomial logistic regression is used to
model nominal outcome variables, in which
the log odds of the outcomes are modeled as
a linear combination of the predictor
variables.
 When the response variable assumes more
than two values not ordinal.
◦ Example: unemployment status: Out of LF,
Unemployed, Employed.
 Sometimes, multinomial logistic regression is
called polytomous logistic regression model.
 It is a natural generalization of the binary
logistic model.

 It is equivalent to loglinear model with


grouped data.

 Statistical software for estimating the model


is widely available.
N.B.: 𝜋𝑘 𝑥𝑖 = 1 − 𝜋𝑗 𝑥𝑖 , 𝑗 = 1, 2, … , 𝑘 − 1
 A positive coefficient of an independent
variable ( 𝑥𝑘 ) implies an increased odds of
observing an observation in category 𝑗 rather
than category 1 “reference category” as 𝑥𝑘
increases, controlling for the effect of other
independent variables.
 A negative coefficient implies that the
chances of being in the reference category
are higher relative to category 𝑗 as 𝑥𝑘
increases.
 A positive coefficient of an independent
variable ( 𝑥1𝑘 ) implies an increased odds of
observing an observation in category 𝑗 rather
than category 1 “reference category” when
being in the category 𝑘 of the variable 𝑥𝑘 if
compared to being in the reference category
of 𝑥𝑘 , controlling for the effect of other
independent variables.
 Study problem: Entering high school students
make program choices among general
program, vocational program and academic
program. Their choice might be modeled
using their writing score and their social
economic status.

 The outcome var. is PROG, the program type,


1 for general, 2 for academic (reference
category) and 3 for vocational.
 The predictors:
 Ses: socioeconomic status of the student.
(categorical var.), 1 for low, 2 for middle and
3 for high.

 Write: writing score of the student.


(continuous variable)
 First, run the needed descriptive stat.
(between the outcome and the predictors).

 Crosstab between prog and ses.

 Descriptive of write at different levels of prog.


𝑝 𝑝𝑟𝑜𝑔 = 𝑔𝑒𝑛𝑒𝑟𝑎𝑙
ln = 1.689 + 1.163 𝑠𝑒𝑠 = 1 + 0.630 𝑠𝑒𝑠 = 2 − 0.058 (𝑤𝑟𝑖𝑡𝑒)
𝑝 𝑝𝑟𝑜𝑔 = 𝑎𝑐𝑎𝑑𝑒𝑚𝑖𝑐

𝑝 𝑝𝑟𝑜𝑔 = 𝑣𝑜𝑐𝑎𝑡𝑖𝑜𝑛𝑎𝑙
ln = 4.236 + 0.983 𝑠𝑒𝑠 = 1 + 1.274 𝑠𝑒𝑠 = 2 − 0.114(𝑤𝑟𝑖𝑡𝑒)
𝑝 𝑝𝑟𝑜𝑔 = 𝑎𝑐𝑎𝑑𝑒𝑚𝑖𝑐
 A one unit increase in the var. write associate
with 0.058 decrease in the relative odds of
being in general versus academic.

 The odds ratio for a one-unit increase in the


variable write is .9437 for being in general
program versus academic program.

 A one unit increase in the var. write associate


with 0.114 decrease in the relative odds of
being in vocational school versus academic.
 The relative odds of being in general versus academic
is increased by 1.163 when moving from the high
level of ses (ref. ses = 3) to a low level of ses (ses =
1).

 The relative risk ratio switching from ses = 3 to 1 is


3.199 for being in general program versus academic
program. In other words, the expected risk of staying
in the general program is higher for subjects who are
low in ses.

 The relative odds of being in vocational versus


academic is increased by 1.274 when moving from
the high level of ses (ref. ses = 3) to a middle level of
ses (ses =2).
 Long, J. S. and Freese, J. (2006) Regression
Models for Categorical and Limited
Dependent Variables Using Stata, Second
Edition. College Station, Texas: Stata Press.

 Hosmer, D. & Lemeshow, S. (2000). Applied


Logistic Regression (Second Edition).New
York: John Wiley & Sons, Inc.

You might also like