Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
96 views2 pages

ME781 Midsem 2016

This document appears to be for a midterm exam covering topics in statistics and regression analysis. It includes 9 questions assessing understanding of nonlinear and linear regression models, k-nearest neighbor regression, accuracy vs precision, forming regression equations with discrete predictors, matrix formulation of multiple linear regression, logistic regression, set operations, and deriving probability density functions based on transformations of random variables.

Uploaded by

Apoorv
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
96 views2 pages

ME781 Midsem 2016

This document appears to be for a midterm exam covering topics in statistics and regression analysis. It includes 9 questions assessing understanding of nonlinear and linear regression models, k-nearest neighbor regression, accuracy vs precision, forming regression equations with discrete predictors, matrix formulation of multiple linear regression, logistic regression, set operations, and deriving probability density functions based on transformations of random variables.

Uploaded by

Apoorv
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/ 2

First Mid Term Examination, ME-781, September 10, 2016

Name: Roll No:


Total Time 2 hours; Total Marks 100

Open notes (self hand-written) examination.


20 1. Let a nonlinear regression model of the type
Y= 0 + 1X +2X2 + 1
approximate the true relation between X and Y then derive an expression for 0, 1
and 2.
If instead of a nonlinear model we assume a linear regression model of the type
Y= 0 + 1X + 
Then, compare the coefficients of the linear and nonlinear model.

10 2. Provider reasoning to show that the kth nearest neighbor regression would perform
very poorly (for a single predictor with the underlying true model being linear) if
a.) the test data is very sparse
b.) the test data is very large but with a large random error (zero mean and not a
function of predictor)

5 3. Provider schematic to show the difference between accuracy and precision and
comment on the role of bias and random error(with zero mean and not a function of
predictor).
5 4. In a linear regression if the predictor has only 3 discrete levels then show how you
would form the equation for your model.
5 5. Multiple linear regression model has the form

With

This can be written in matrix form as:

𝑌 = 𝑋𝛽
Where, Y, X and  are matrix of the size nx1, nx(p+1) and (p+1)x1, respectively. (Note
that n is the number of training data points, and p is the number of predictors)
And
𝑅𝑆𝑆(𝛽) = (𝑌 − 𝑋𝛽)𝑇 (𝑌 − 𝑋𝛽)
And
𝜕𝑅𝑆𝑆(𝛽)
= −2𝑋 𝑇 (𝑌 − 𝑋𝛽)
𝜕𝛽
Show that the choice of  which minimizes the RSS leads to residual vector (𝑌 − 𝑋𝛽)
becoming orthogonal to column space of X.
5 6. Probability of a grad student owning a car is dependent upon his or her graduate salary.
Let this relation be modeled by Logistic regression with logistic function as

. If the model parameters 0 and 1 what are -10 and 0.006, then
what are the odds that a grad student having a salary of 3000 owns a car.
25 7. Let set A=[1,3]x[3,6] and set B=b((3,3),1). Please note B is a closed ball at (3,3) of radius
1.
Then draw the following:
a.) A U B
b.) A ∩ B
c.) Acl - Ain
d.) A ⊕ B
e.) A ⊖ B
10 8. X is a uniform random variable in [0,1] and Y = sin-1(X). Y is defined in [0,Pi/2]. Then find
the probability density function of Y.
15 9. Let Y = a|X| + b, where X is a random variable. Derive an expression for probability
density function of Y.

You might also like