Scribd
Upload
21 views22 pages

Transforming From Exponential To Linear Growth - Video

The document discusses the process of transforming exponential growth data into linear growth for analysis using logarithmic regression. It explains how to implement this transformation in Excel, including taking the logarithm of the dependent variable and using it in a linear regression model. Additionally, it covers how to generate forecasts based on the regression output and the interpretation of the resulting coefficients.

Uploaded by

Aghachakla Gologoldoon
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
21 views22 pages

Transforming From Exponential To Linear Growth - Video

The document discusses the process of transforming exponential growth data into linear growth for analysis using logarithmic regression. It explains how to implement this transformation in Excel, including taking the logarithm of the dependent variable and using it in a linear regression model. Additionally, it covers how to generate forecasts based on the regression output and the interpretation of the resulting coefficients.

Uploaded by

Aghachakla Gologoldoon
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 22

Transforming from

Exponential to
Linear Growth
Graph of Data vs Time
2500

What if your 2000

data is
growing
1500

exponentiall 1000

y?
500

0
0 2 4 6 8 10 12 14
Graph of Data vs Time
2500

Q: How do 2000

we use
linear 1500

regression
on this data 1000

set? 500

0
0 2 4 6 8 10 12 14
Graph of Data vs Time
2500

2000

A: We take
a logarithm 1500

of the y-
values… 1000

500

0
0 2 4 6 8 10 12 14
Graph of Data vs Time
2500

A: Which 2000

‘dampens’
the growth 1500

back down
to linear 1000

growth… 500

0
0 2 4 6 8 10 12 14
Graph of Data vs Time
14

A: Which 12

‘dampens’ 10

the growth 8

back down 6
to linear
growth… 4

0
0 2 4 6 8 10 12 14
Graph of Data vs Time
14

So that we
can use
12

linear 10

regression 8

to forecast 6

with our 4

data 2

0
0 2 4 6 8 10 12 14
Graph of Data vs Time
14

12

So how do 10

we do 8

this??? 6

0
0 2 4 6 8 10 12 14
• We take the dependent variable
(or, take the data values if we
only have data & time as our
variables)
So how do • We take a log of the dependent
we do variable
this??? • ie: =
• We use the logged variable as
the dependent variable in our
linear regression (multiple
regression) model:
• We take the dependent variable
(or, take the data values if we
only have data & time as our
And how do variables)
we • We take a log of the dependent
implement variable
this in • ie: =
Excel? • We use the logged variable as
the dependent variable in our
linear regression (multiple
regression) model:
• Insert a new column in Excel between
the independent variables & the
dependent variable Insert a new
column in Excel
And how do
• In that column, calculate the log of the
we dependent (data) variable using
implement ‘=log()’ in Excel
this in • Use that logged variable as the
Excel? dependent variable and use the
original x variable(s) as the
independent variables when using the
Data analysis Toolpak’s linear
regression
Here is our
output in
Excel from
our
example…
How do we
generate
forecasts
based off of
this output?
…
• We use the coefficients given
in the regression output to
How do we generate the following
generate equation:
forecasts
based off of • Now how do we generate
this output? forecasts based off of this
… equation?
• Let’s go back to the “general”
equation which is:
• We need to ‘undo’ what we just
did – which was to log the y
variable
How do we
• We ‘undo’ a logarithm by taking
generate the inverse:
forecasts The inverse of is
based off of • This gives the following equation:
this output?
…
Note: we put both side of the
equation to the power of 10
• Let’s “clean up” the equation that
we just generated on the last slide:
How do we • The log and the base 10 will cancel
generate each other out on the left-hand side:
forecasts
based off of • We can then use the following
this output? exponent rule on the right-hand
… side:

• This gives:
• Let’s “clean up” the expression below:

How do we • Let’s assign variables to the powers of 10:


generate
forecasts
based off of • This transforms our forecast equation:
this output?
… • Note: you don’t need to assign those
variables if you don’t wish – it’s fairly
standard practice – so I show you that
way!
• In our example, we generated
the following forecast equation:

Where and
Now, back
to our • This gives the following values
for b :
example…

• Then our forecast equation


becomes:
• Finally, let’s forecast the
value of y when x = 12 using
the following equation:
Now, back
to our Plugging in x = 12 gives:
example…
• Which gives the following
error:
• Let’s analyse the meaning of the
coefficients in forecast equation:

Let’s  = the y-intercept (or the estimated initial


value in period if we are using coded time
analyse the periods for our independent variable).
results from  In our example this means: the estimated
our starting value of our investment is $93.04
 helps give us the estimated annual
example… compound growth rate (as a percentage):
 Estimate growth rate .
 In our example this means: the estimated
Estimate growth rate
Note: the data in this example
was generated by the following:
 Assume we start with $100 in an
Generating investment
the ‘data’  Assume the invest earns, on
for our average, 24% per year
example…  But it doesn’t always earn exactly
24% per year. Let’s say the
standard deviation of those
earnings is 5% and assuming those
rates are normally distributed…
Note: the data in this example was
generated by the following:
 This means
 sometimes, the rate could be as low as
Generating 24 – 3 x 5 (3 st dev’s below the average) =
9%
the ‘data’  sometimes, the rate could be as high as
for our 24 – 3 x 5 (3 st dev’s above the average) =
39%
example…  We can simulate this rate using:
 = norm.inv(rand(), 24, 5)
 The above equation is called Monte Carlo
simulation & we will see it later in the
course

You might also like