# Applied Managerial Decision-Making: Regression Models

The purpose of this post is to explain to the manager of Widge Corp. about what steps need to be taken in order to be able to branch out using the cold beverages that are offered at the company. I will be explaining what type of regression model will and can be used in order to make this a good transition and help to increase our monthly forecast of sales preferably in the next year. It will also include the different variables will be a part of the projected model. The first thing that needs to occur in this process is to explain what it is that regression model are and what they do.

There are in fact about four different types of regression models that can be used when trying to figure out this process. They are: simple linear regression, multiple regression, linear regression, and non-linear regression. Simple linear regression intels establishing a relationship between selected and observed values which are constructed so that there is a more probable value of an observed value which can be predicted from the values of a selected variable. This is variables that are observed are that of x (selected variable) and y (observed variable). An independent variable is what the selected variable is essentially called as well as being a predictor, the carrier, or even the input per se. Then again, when we talk about the observed value/dependent variable, which is then seen as being the output or the response. If we look at Company W. and its sales regarding cold beverages which are seen as being dependent upon its marketing budget, the bigger the budget for marketing is it will translate into a bigger increase of sales for the cold beverages. When looking at it from a regression standpoint, it would make for the sales being a regression on the marketing budget (ReliaSoft Corp., 2008).

The next type of regression model is the multiple regression where relationship depends on two or more variables that are independent of each other. Then the value that is observed is made so that the most detrimental value of the observed value will be thought of from the values that were of the selected variables. With this analysis in which there are numerous variables which have an influence on a dependent variable, it takes a multiple regression of the form of y = b (1) x (1) x b(2) x (2) + _ + b(n) x (n) + c. To elaborate on this further, if we look at c, it is a constant per se. Then as far as the b, it makes for the coefficient of regression which stands for the amount in which the variable y changes when it matches an independent variable and it changes as a result of 1. A multiple correlation coefficient is what represents the variance percentage in the dependent variable that explains all of the collective variables. This is seen as being R squared. Now if Widge Corp. makes a conclusive decision that the sales of cold beverages actually depends on the actual marketing budget, then the amount of distributors that will be appointed will in essence have to look at the beverages actual prices and the different types as well. From a multiple regression standpoint, a number of assumptions can be made which will most likely say that there is a correlation and linearity of relationships in question. It will also look at the similarity throughout the independent variables ranges and its complete data range with the absence of any outliers (Palmer, 2008).

From the linear regression perspective, this process looks at a straight line which focuses on a set of data which will look for future outcomes. This straight line is in fact used to determine what effect an independent variable has on a specific variable. In saying this, the relationship of variables from a linear standpoint we can say when it comes to Company W, we can figure out the relationship between the sales of cold beverages and the marketing budget, we could say that there are variables (which are x and y) and the linear regression attempts to make for this relationship fit on a straight line to actually fit certain data. With this regression model, it is said that y = a + bx + e. The e is fact labeled as being a residual number (Rao, S., 2003).

To talk about the fourth type of regression model, we will look at the non-linear model. It basically is the relationship of more than one selected variable and an observed variable in a non-linear way. This model is seen as when model y = f(x,t) + e in which it is multi-dimensionally applied to x and y data where the f is considered non-linear with not being able to detect the parameters (which are seen as t). In Company W’s case, they should get the parametric values which are a part of the least squares typically saying. The function for this could be something such as: f(x) = ax(squared) + bx + c. Problems occur though with linear regression in that it can be linearized by using a certain model of transformation (Harrell, no date).

I have in the previous paragraphs, gone over the different types of regression models that Company W might be able to use in their quest for their wanting to come up with a model to forecast monthly sales of cold beverages for the next year. Although this type of thing can be used in many different areas, it can just as well help to determine the probability of just about anything relating to business just to name one. Widge Corp and its beverage sales can actually be predicted just by looking at the marketing budget that the company has set, I addition to its number of people appointed, and how many ad campaigns it has launched over the tenure. Other things to look at are things such as: areas like PR, its sales persons, and not forgetting to mention its different flavors that the company already has and want to launch. It can actually change the variables at different times to figure out which way sales the best. Once they find the most suitable way, then they will be able to select any combination of things which meet their objectives from a cost effective standpoint.

References:

Harrell, F., (no date). Regression Model Strategies. Retrieved on January 30, 2012 from http://www.stat.ucla.edu/~rgould/asw2005/abstract.pdf Palmer, M., (2008). Multiple Regression. Retrieved on January 29, 2012 from http://ordination.okstate.edu/MULTIPLE.htm Rao, S., (2003). Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. Retrieved on January 29, 2012 from http://www.questia.com/googleScholar.qst?docId=5002529587

ReliaSoft Corporation, (2008). Simple Linear Regression Analysis. Retrieved on January 29, 2012 from http://www.weibull.com/DOEWeb/simple_linear_regression_analysis.htm