You are on page 1of 3

Given the expression

at + b t at
yt = =1+ (1)
bt bt
the change of yt for the time t = [0, 1] may be written
   
a1 a0 a1 a0
t = y1 y0 = 1 + 1+ = (2)
b1 b0 b1 b0

It is now of interest to investigate how much the respective change in a = a1 a0 and


b = b1 b0 has affected the total change y.

First calculating the shared denominator


a1 b 0 a0 b 1
y = (3)
b0 b1
then adding and subtracting a1 b1 yields
a1 b 0 a0 b 1 a1 b0 a1 b1 + a1 b1 a0 b1
y = = (4)
b0 b1 b0 b1
This allows for setting coefficients outside the parenthesis of a and b respectively
a1 1 1 a1
y = (b0 b1 ) + (a1 a0 ) = a b = c1 a c2 b (5)
b0 b1 b0 b0 b0 b1
1 a1
where c1 = b0
and c2 = b0 b1
.

In general this is a function of the form

f (x, y) = c1 x c2 y (6)

i.e. a linear function of two variables, x and y, with the coefficients (c1 , c2 ) R. The
change f (x, y) experiences as a function of x would then be

f (x, y)
= c1 (7)
x
and for y
f (x, y)
= c2 (8)
y
And a change in f (x, y) would be

f (x, y) = f (x, y) = c1 x c2 y (9)

1
the proportional change that x and y has accounted for is just the fraction of the
coefficients, i.e. the influence of x is
c1
(10)
c1 + c2
and the influence of y
c2
(11)
c1 + c2
where both fractions can be multiplied with 100 to get it in percentages.

However coming back to the specific case at hand where the considered equation is
a1 1 1 a1
y = (b0 b1 ) + (a1 a0 ) = a b = c1 a c2 b (12)
b0 b1 b0 b0 b0 b1
it is not quite so trivial, because c1 and c2 are functions of a1 , b0 and b1 and hence when
calculating the derivatives
y
6= c1 (13)
a
and
y
6= c2 (14)
b
however an attempt can be made to analyse the expression.

It is a bit difficult to explain, but there are two different ways to look at the influence
of a and b on y. In mathematical terms, the derivatives would be the appropriate
y
function to consider, i.e. a and y
b
which would not be directly possible. However
merely considering the derivatives would not account for the fact that a and b is
also influencing. If it is assumed we are at t = 1, meaning a0 , b0 , a1 and b1 are known,
then neither c1 , c2 nor a and b are really variables. Hence the proportional change
is actually found as follows:
For a
c1 a
(15)
c1 a + c2 b
and for b
c1 b
(16)
c1 a + c2 b
where both can be multiplied by 100 to get it in percentage. These satisfy the constrains
that
c1 a c1 b
+ =1
c1 a + c2 b c1 a + c2 b
c1 a
= 0, for a = 0 (17)
c1 a + c2 b
c1 b
= 0, for b = 0
c1 a + c2 b

2
which is the minimum requirement for it to be true. However I am not completely sure
that (15) and (16) are true, as I have no solutions I know to be true, to compare them
to.