Help a programmer with very simple Linear Regression (ordinary least squares?)

#1
Hi everyone,

I'm a programmer, so I don't have much understanding of statistics. But I am supposed to write a program for my company that predicts our inventory needs for the purchase orders, and my guess at the best model to use is the ordinary least squares regression line. I may need to use the logarithmic trend line instead, but you're the statistics experts, so use your best judgment.

Problem: I have no clue how to work out the formula... I'm going to have a hard enough time trying to convert the formula to an SQL statement, but that's another story.

Here's the data I need to figure out: I have to examine the percentage of increase number of a product sold for the past 4 months. I'd also like to average those increases with the corresponding month from the previous year. That should give me 3 numbers to graph as points where x=month and y=quantity sold.

I then need to use the linear regression to predict the quantity I need to order for the next 3 months. I can do ALL of this in Excel, but I need the formula to make this work in my SQL program. In fact, I've attached a screenshot of the graphical representation of the formula I'm trying to create.

Please, help a guy out!
 
#2
My graphical representation

Sorry, forgot to attach the graph I did in Excel. This graph is actually a Logarithmic trend line.
 
Last edited:
#3
all statistical software packages have implemented functions to perform linear regression. you don't need to write your own function.
 

mp83

TS Contributor
#5
Just get your data from your SQL or whatever database into Excel and use LINEST() function [link] or the Data Analysis Toolpak [link ; link2]

If you consider more advanced modelling you can realte your database to R.