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

Logistic Regression

Logistic regression predicts categorical outcomes like binomial events (y = 0 or 1). It models the data using the logistic function to calculate the probability of an event occurring given input variables x. The key parameters estimated are the coefficients B0, B1, etc. which represent the log-odds of a one unit change in the associated input feature. These coefficients are used to calculate the log-odds, odds, and probability of the event based on the feature values. For example, analyzing student honors data found the odds of a female student graduating with honors was 82% higher than a male student.

Uploaded by

shahid
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
94 views

Logistic Regression

Logistic regression predicts categorical outcomes like binomial events (y = 0 or 1). It models the data using the logistic function to calculate the probability of an event occurring given input variables x. The key parameters estimated are the coefficients B0, B1, etc. which represent the log-odds of a one unit change in the associated input feature. These coefficients are used to calculate the log-odds, odds, and probability of the event based on the feature values. For example, analyzing student honors data found the odds of a female student graduating with honors was 82% higher than a male student.

Uploaded by

shahid
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 12

Logistic Regression

Machine Learning

Lecture Slides
Introduction
• A popular statistical technique to predict binomial outcomes (y = 0 or 1) is
Logistic Regression. Logistic regression predicts categorical outcomes
(binomial / multinomial values of y), whereas linear Regression is good for
predicting continuous-valued outcomes (such as weight of a person in kg, the
amount of rainfall in cm).
• The predictions of Logistic Regression (henceforth, LogR in this article) are in
the form of probabilities of an event occurring, ie the probability of y=1,
given certain values of input variables x. Thus, the results of LogR range
between 0-1.
• LogR models the data points using the standard logistic function, which is an
S- shaped curve given by the equation:
 Concepts

• The equation to be solved in LogR is:

• where:
• p = probability that y=1 given the values of the input features, x.
• x1,x2,..,xk = set of input features, x.
• B0,B1,..,Bk = parameter values to be estimated via the maximum likelihood method. B0,B1,..Bk
are estimated as the ‘log-odds’ of a unit change in the input feature it is associated with.
• Bt = vector of coefficients
• X = vector of input features
• Estimating the values of B0,B1,..,Bk involves the concepts of probability, odds and log odds. Let
us note their ranges first:
• Probability ranges from 0 to 1
• Odds range from 0 to ∞
• Log odds range from -∞ to +∞
Example
The task is to predict which students graduated with honours or not (y = 1 or 0), for 200
students with fields female, read, write, math, hon, femalexmath . The fields describe the
gender (female=1 if female), reading scores, writing scores, math scores, honours status
(hon=1 if graduated with honours) and femalexmath showing the math score if female=1
.
The crosstab of the variable hon with female shows that there are 109 males and 91
females; 32 of those 109 females secured honours.

Probability:
The probability of an event is the number of instances of that event divided by the total number of
instances present.
Thus, the probability of females securing honours:

= 32/ 109
= 0.29
Odds:
• The odds of an event is the probability of that event occurring
(probability that y=1), divided by the probability that it does not occur.
• Thus, the odds of females securing honours:

= 32/77
= 0.4155
≈ 0.42
This is interpreted as:
1. 32/77 => For every 32 females that secure honours, there are 77 females that do not
secure honours.
2. 32/77 => There are 32 females that secure honours, for every 109 (i.e 32+77) females.
Log odds:

The Logit or log-odds of an event is the log of the odds. This refers to the
natural log (base ‘e’). Thus,

Thus, the log-odds of females securing honours:


Odds ratio:
It is the ratio of 2 odds; these 2 odds are obtained at 2 different values of x, the 2 values of
x being 1 unit apart.
Such as: The odds obtained when x=0 and x=1 (ie when there is a 1 unit change in the value
of x, where x=0 denotes male and x=1 denotes female).
Q: Find the odds ratio of graduating with honours for females and males.

=> As OR= 1.82, the odds for


females graduating with honours
is about 82% higher than the
odds for males graduating with
honours.
Calculations for probability:
Suppose we want to calculate the effect of being female on the probability of
graduating with honours.

Where :
1. B0,B1,..Bk are estimated as the ‘log-odds’ of a unit change in the input feature it is
associated with.
2. As B0 is the coefficient not associated with any input feature, B0= log-odds of the
reference variable, x=0 (ie x=male). ie Here, B0= log[odds(male graduating with
honours)]
3. As B1 is the coefficient of the input feature ‘female’,
1. B1= log-odds obtained with a unit change in x= female.
2. B1= log-odds obtained when x=female and x=male.
Calculations:
• From the calculation in the section ‘odds ratio(OR)’,
• B1= log (1.82)
• B1= 0.593
• Thus, the LogR equation becomes
• y= -1.47 + 0.593* female
• where the value of female is substituted as 0 or 1 for male and female respectively.
• Now, let us try to find out the probability of a female securing honours when there
is only 1 input feature present-‘female’.
• Substitute female=1 in: y= -1.47 + 0.593* female
• Thus, y=log[odds(female)]= -1.47 + 0.593*1 = -0.877
• As log-odds = -0.877.
• Thus, odds= e^ (Bt.X)= e^ (-0.877)= 0.416
• And, probability is calculated as:
Thus, the probability of a female
securing honours when there is
only 1 input feature
present-‘female’, is 0.29.

Thus, methodology of LogR:


1. To find the values of coefficents B0, B1, B2,…Bk to plug into the equation:
y= log(p/(1-p))= β0 + β1*x1 + … + βk*xk = Bt.X , for specific values of x.
2. The result of substituting the values of B0,B1,B2,..Bk and the values of x into this
equation, is the log-odds of an event (ie log-odds of y=1, given those values of x). Thus,
the coefficients are obtained in the log scale.
3. Now, convert the coefficients into the odds scale and later into probability. Using the value
of log-odds (Bt.X) of the event, odds is obtained by e^ (Bt.X). Then, the probability of the
event is derived by {e^ (Bt.X)/[1+e^ (Bt.X)] }.

You might also like