Are You Trendy ?
Chandoo is off Holidaying teaching excel in the Maldives and has lent me the keys to his Blog (Chandoo.org) and this week I plan to take it for a spin.
I will be posting 3 posts on Trend Analysis/Forecasting using Excel and a forth post on some Hidden Worksheet Properties which I stumbled onto last week !
Hopefully if I look after the Blog while Chandoo is gone, He will let me borrow the keys another day.
“Tomorrows weather will be fine and hot with a chance of showers in the morning.”
We have all seen this type of forecasting during the nightly news.
This week I am going to go through the basics of forecasting and trend analysis using Excel as a tool.
We will look at some simple trends and make predictions about future values.
In later posts we will look at more complex data and other methods of tackling these analysis.
Often you may have a set of data and need to know what an intermediate or future value of that data may be.
This week we will investigate 3 methods of tackling this problem using Excel.
In this post we’ll look at doing forecasting manually
In the second post we’ll look at a few excel functions that assist us with forecasting
The third post will discuss a method of looking at any value along an Excel generated Trend Line and give you a tool to assist you in this.
In all environments where numbers are collected and people make use of these numbers the ability to forecast or extrapolate data may be required.
In forecasting we are going to look at the trends that the data has and use these trends to help forecast future values or values outside the measured data. The trends can also be used to infill data where gaps may be missing in the collected data.
This post will look at doing this manually, albeit with some help from Excel.
We will examine a business that makes things and we will measure some measurement of those things every 5 days. In trend analysis it doesn’t matter what you measure or what your measuring it against.
We have collected some data which is tabulated
One of the easiest ways to visualise this relationship is to draw a quick chart of one measure vs a base or in our case a time line.
This can be shown graphically as a simple Excel Scatter chart
You can see that there is some level of variability in the measurement as the data doesn’t quite fit a straight line.
Manually we can make an estimate of a line of best fit and draw it on the chart by adding a new data series consisting of 2 points.
There are 3 quick methods of using this line of best fit
- Manual Estimates
- Equal Triangles
- Equation for the line
If we want to know what the measurement would be for a location where no measurement was taken we can use the chart and 2 quick lines to show in this example that for 20 days we would expect a measurement of about 26 units.
This can also be used for extrapolation of our data past the limits of what was measured.
By extrapolating the Line of Best Fit beyond the data, the same technique can be applied to estimating what some future value maybe.
Equal Triangles is a technique where a simple ratios of 2 similarly shaped but different sized right angle triangles can be used to make estimates of missing or extrapolated data.
Using Equal Triangles the ratio of the height to the width of Triangle 1 (Red) is equal to the ratio of the height to the width of Triangle 2 (Blue).
So in the example above
Y1/X1 = Y2/X2
Y1 = 38 – 8 = 30
X1 = 30 – 5 = 25
Y1/X1 = 30/25 = 1.2
So for Triangle 2
Y2/X2 = 1.2
Y2 = ? – 8
X2 = 20 – 5 = 15
from Y2/X2 = 1.2
(? – 8 ) /15 – 1.2
We can rewrite this as
? = 8 + 1.2 x 15 = 26.0
Unknown Y = Min Y + Ratio x (New X – Min x)
Once we have an equation we can setup a new series on out chart based on an equation in some cells and then directly plot the data onto our chart.
In this case we have used the equation =F105+1.2*(E111-E105)
Equation of the line of Best Fit
If we are using a straight line to model our line of best fit, we can also write an equation for the line in the form
Y = mX + c
Where: Y is the unknown measure
X is the X value for which we want to know the value of Y
m is the gradient of the line
c is the Y intercept of the line (or Y value when there is no X value or X =0 )
The gradient m is calculated as the Rise / Run or in our example 30/25 = 1.2
The Y Intercept is the value when x = 0. This can be back calculated from the first point (5,8)
C = 8 – (5 x 1.2) = 2.0
So the equation for our line of best fit is Y = 1.2 X + 2
We have used this in the next example =E136*1.2 + 2
The good thing about having an equation for the line is that we can use that to calculate any value of our measure.
So if we want to know the measure on a day outside the range we measured, say the 40th day
You can download examples of all the above charts from the following link
Benefits of Manual Estimation
- Applicable to simple models
- Can be used without a computer or a calculator in the field
- Gives the user a better feel for the data
Problems of Manual Estimation
- Only applicable to simple models
- Reliant on the accuracy of your estimate of the trend
- No measure of how accurately your estimate fits the data
In the next post we will look at using excel functions to automatically estimate lines of best fit and other excel functions to aid in estimation of non-linear functions.