# Thread: Which Statistical Analysis Method Should Be Used To Solve Forecasting?

1. ## Which Statistical Analysis Method Should Be Used To Solve Forecasting?

Hello,

I am trying to forecast/predict how many confirmed sales will occur within the next 6 months. Sales are confirmed via computer (online) or by a person (in-store). The data set below displays how many actual orders were confirmed during May-November of 2017.

 Month Computer Person Totals May 3850 546 4396 June 3801 807 4608 July 3596 599 4195 August 3751 732 4483 September 3485 642 4127 October 3447 584 4031 November 3720 594 4314 Totals 25650 4504 30154

Any help with this is much appreciated!

2. ## Re: Which Statistical Analysis Method Should Be Used To Solve Forecasting?

Hey mathquestions321.

You will need to have some assumptions and then use that to translate what you have into a mathematical statistical model.

What sort of things are you going to assume?

3. ## Re: Which Statistical Analysis Method Should Be Used To Solve Forecasting?

A "least squares regression" line would probably be best to estimate future sales. What software/hardware do you have? The TI-84 has a built in regression program- here's a tutorial on using it: TI-84: Least Squares Regression Line (LSRL) | TI-84 Graphing Calculator | CPM Student Tutorials.

If you have "Excel" look at Regression in Excel.

4. ## Re: Which Statistical Analysis Method Should Be Used To Solve Forecasting?

Chiro - I'm not sure what you mean by assumptions? Can you please elaborate for me a little bit.

HallsofIvy - I am using Excel for my statistical analysis. I know regression is used to determine the relationship between a dependent and independent variable. I guess I am a bit lost on how that method would work in this case?

5. ## Re: Which Statistical Analysis Method Should Be Used To Solve Forecasting?

Linear regression gives you a line, with the sales as a function of the month, that comes closest (in the least squares sense) to the given data points. Here I would expect that you use May= 0, June= 1, etc. to November= 6. Then evaluate that function for months 7 to 12, the next six months.

6. ## Re: Which Statistical Analysis Method Should Be Used To Solve Forecasting?

Well usually when you have data you need to apply further conditions like whether a model exists [like the relationships between the variables and some statistical constraints] along with a reason of some sort for why you pick them.

Just to think about a simple example - for your model imagine a linear model Totals = Computer(Month) + Person(Month) + e as some random variable.

There are reasons for why you can limit one variable to be say Poisson [if it's got to do with a rate] or say Binomial [if it's some kind of count between a fixed number of integers] or even a continuous variable versus a discrete one.

For Checking computer and person sales - if they are independent then you don't have a co-variance term and this will simplify the model.

I shouldn't do this problem for you but you should consider what relationships exists with Person and computer counts as a function of month and use that to get a model for totals as a function for month which gives you estimates on the parameters so that you plot your expected totals as a function of month and get an estimate based on that at what it should be.