IE 202 Study Set 4 Fall 2020

Shadow Price in Linear Programming: A shadow price or a dual price is associated with each
constraint of the model. It is the optimal value of the dual variable corresponding to this constraint.
Then the shadow price associated with a particular constraint tells you how much the optimal value
of the objective would change per unit change in the associated constraint, assuming the basis remains
optimal. It will increase (decrease) if the objective is maximization (minimization) and the restriction
becomes loose (increase for ≤ and decrease for ≥).
Reduced Cost in Linear Programming: Reduced cost vector = c–cB B −1 A, i.e., the negative of
the values in row 0 of the tableau. The reduced cost value of a nonbasic variable indicates how much the
objective function will worsen per unit increase in the value of the associated variable. It also tells us how
much the coefficient on the corresponding variable must be changed (increased or decreased, whichever
is better), before the value of the variable will be positive in the optimal solution.

Q1: Consider the following problem American ale company. The company can contract with brewery
1 at 100 dollars per barrel and brewery 2 at 125 dollars per barrel. The company promised a
production of 10000 barrels of ale to its customers within the existing planning horizon. The
production requirements necessitate that the number of barrels purchases from brewery 1 should be
at least that from brewery 2. However, brewery 2 purchases should be at least a quarter of brewery
1 purchases. The following LP (decision variables and the model are self explanatory) solves this
company’s problem.

min 100Q1 + 125Q2

s.t. Q1 + Q2 ≥ 10000
−0.25Q1 + Q2 ≥ 0
Q1 − Q2 ≥ 0
Q1 , Q2 ≥ 0
The Excel Solver sensitivity report for the LP model is given below:

Answer the following using the Excel output.

1
 (a) What is the optimal solution and the optimal solution value?
(b) Which constraints are binding?
(c) Determine the range of objective coefficients such that the current optimal solution remains
optimal.
(d) What would happen to the objective function if the right hand side of the first constraint is
increased by 1 unit?
(e) What would happen to the objective function if the right hand side of the second constraint
is increased by 1 unit?
(f) What would happen to the objective function if the right hand side of the third constraint is
increased by 1 unit?

Q2: Consider the LP

max 100B + 300M + 50C

s.t. 300B + 1200M + 120C ≤ 93000
0.5B + M + 0.5C ≤ 101
B , M , C ≥ 0

where P is the profit of ordering B units of bicycle, M units of Mopeds and C units of child seats.
First constraint ensures that the budget is not exceeded while second constraint ensures that storage
capacity is respected. The optimal solution can be found below:

The Excel Solver sensitivity report for the LP model is given below:

2
 Answer the following using the Excel output.

(a) What is the optimal solution and the optimal solution value?
(b) Which constraints are binding?
(c) Determine the range of objective coefficients such that the current optimal solution remains
optimal.
(d) Determine the range of right hand side values such that the current basis remains optimal.
(e) What would happen to the objective function if the right hand side of the first constraint is
increased by 1 unit?
(f) What would happen to the objective function if the right hand side of the second constraint
is increased by 1 unit?
(g) Using the Excel output only, construct Row 0 and Rhs values in the optimal table.

Q3: Consider the following LP and the corresponding optimal tableau for the next questions.

max 2x1 + x2 − x3
s.t. x1 + 2x2 + x3 ≤ 8
−x1 + x2 − x3 ≤ 4
x1 , x2 , x3 ≥ 0

Basis z x1 x2 x3 x4 x5 RHS
z 1 0 3 3 2 0 16
x1 0 1 2 1 1 0 8
x5 0 0 3 -1 1 1 12

(a) Find optimal basis B, its inverse B − 1,cB and the optimal dual vector ω.
(b) Suppose that the right hand side of the first constraint is changed from 8 to 8-4 and the right
hand side of the second constraint is changed from 4 to 4 + 24. For which values of 4 does
the current basis remain optimal? How does the optimal bfs and the optimal value change
with respect to 4?

3
 (c) Add the constraint x1 + 3x2 + x3 ≥ 9 to the above LP. Is the current solution still optimal? If
not, find (if any) the new optimal solution using sensitivity analysis.

Q4: Consider the following LP:

min x1 + 2x2
s.t. x1 + 3x2 ≥ -1
x1 + x2 ≥ 1
x1 ≤ 2
x1 , x2 ≥ 0

You know that the optimal solution is (x1 , x2 ) = (1, 0).

(a) Convert the LP to standard form.

(b) Construct the optimal tableau corresponding to the the standard form LP of part (a) without
any simplex opeartions.
(c) Construct the dual of the given LP as well as the dual of the corresponding standard form LP.
What do you notice?
(d) Find the optimal solution of the dual LP in (c) using each of the following three methods:
(i) Using Strong Duality Theorem
(ii) Using Complementary Slackness Theorem
(iii) Reading off from tableau

Q5: Consider the following LP, and its optimal solution. s1 and s2 are the slack variables that were
added to the first and second constraints, respectively.

max 3x1 − x2 + x3
s.t. x1 − 2x2 ≤ 6
x1 + x3 ≤ 10
x1 , x2 , x3 ≥ 0

The optimal tableau for this linear program is as follows:

z x1 x2 x3 s1 s2 RHS
3 1 5
1 0 0 2 2 2 28
x1 1 0 1 0 1 10
1
x2 0 1 2 − 12 1
2 2

(a) Does the current optimal solution remain optimal if the column a3 is changed from (0, 1)>
to (−6, 1)> ? If not, update the tableau and indicate entering and leaving variables. Do not
perform any simplex operations.
(b) Suppose that the cost coefficient vector is changed from c = (3, −1, 1)> to c = (∆, −1, 1)> .
Find the values of ∆ for which the current basis remains optimal.

4
 (c) If you were to choose between increasing the RHS of the first constraint, i.e., b1 by 1 unit and
increasing the RHS of the second constraint, i.e., b2 by 1 unit, which one would you choose
and why? What is the effect of this change on the optimal value of the objective function?
(d) Does the current optimal solution remain optimal if the following constraint is added to the
LP?
x1 + x2 + x3 ≤ 9.

If not, update the tableau and indicate entering and leaving variables. Do not perform any
simplex operations.
(e) Decide if the following statement is TRUE or FALSE. Do not just provide an answer, provide
an explanation along with it.
The reduced cost for a basic variable is zero.

Q6: (a) If an LP in standard maximization form is unbounded, then it is possible to make it have a
finite optimal solution value by changing the right hand side vector?
(b) If an LP in standard maximization form has a finite optimal solution, then it is possible to
make this LP unbounded by changing some parameters? If so which parameters?