You don’t need psychic powers to predict with regression analysis: You simply need the right technique and skills.
According to the University of Chicago, “Regression analysis is a statistical tool for the investigation of relationships between variables.”
What does this mean to YOU?
As the title suggests, you need not to be a psychic or have psychic powers to predict the future. All you need is the right tool and method to make those predictions. In my other articles I have talked about at length on various powerful methods and techniques to perform predictions like:
- Design of Experiments (Design of Experiments)
- Naïve Bayes
- Monte Carlo Simulation
- Benford’s Law (The law of the first digit)
- Association Rules
In contrary to the psychic abilities, statistical predictions are more technique and concept oriented (not that psychic abilities doesn’t have a concept or a technique, we will leave this for our future discussions).
Statistically significant predictions uses mathematical concepts and techniques to prove the relationship of one or more variables with the variable that needs to be predicted. The accuracy of the model depends on how well the data fits in the model.
In this article you will learn how to use regression analysis to perform basic statistical predictions to accelerate your ROI.
The Basic Workflow of Regression Analysis
Prediction with Regression Analysis
The task is to predict how the yearly advertising cost/budget is impacting and explaining the yearly sales volume. This prediction can have multiple independent variables. However, to keep this process simple I have confined this technique to only one independent and one dependent variable. This analysis will help any marketing based organization to allocate appropriate resources efficiently to achieve high sales rates & low operation costs Click & Tweet! .
Let’s Get Started
I used Minitab Statistical software to perform this analysis. However, the same analysis can be performed using Excel which also gives an equation output that can be used to predict sales outcomes based on the amount spent on advertisements.
In the best interest of this article I have already isolated the most impactful variable and would like to leave the variable identification and model trimming process discussion for future articles. During the model trimming process I ensured that there are no variables with multicollinearity (this can be achieved by looking at the VIF scores i.e. Variation Inflation Factors).
Step 1: Load the data
Open Minitab (I am using version 17) and paste the data on to the Minitab Worksheet
Step 2: Perform the Regression Analysis
On the Minitab Toolbar click the Assistant Tab and select Regression
Step 3: Make the choice
Choose appropriate Regression Analysis from the options. For this analysis I used Simple Regression
Step 4: Input Variables “Y” is the function of “X”
Input Advertisement Cost “X” variable field and Sales Unit into “Y” variable field, leave everything the way it is and hit “ok”
Now that I have one predictor Advertisement Budget and one response Sales Unit appropriately placed into the prompts, let’s look at the result.
What does the Regression Analysis tells?This model shows that there is a statistically significant relationship in between the Y (Sales) & X (Advertising Budget or Cost) variables. Click & Tweet! It also tells that what percent of the variation in Sales Unit is accounted by the regression model.
The main question to be answered is “Can I Predict”? And the answer to this question is “yes”. And this is because the fitted equation for this prediction model describes the relationship between Y and X is:
Y = – 36615 + 11.51 X – 0.000201 X^2
As long as the model fits the data well, this equation can be used to predict Unit Sales for a value of Advertisement Cost or Budget.
Similar analysis can be performed using the Excel Spreadsheet that you can download at the end of this article.
Conclusion on Regression Analysis
The above example clearly shows that with, you don’t require psychic powers to predict, you just need a good model with statistically significant relationship and always remember “All models are wrong but some are useful: George Box”
Download Simple Regression Excel Sheet Here