Problems with the Forecasting Chapter

Status
Not open for further replies.
J

Jamie Morris

Steve Prevette said:
We went over the "Forecasting" chapter, which makes heavy use of moving averages and regression. I would also like you to review the attached paper and provide any comments or real life sagas with such misuse of statistical principles.

See http://elsmar.com/Forums/showthread.php?t=9081 for the paper.

Worth +1 point extra credit on the final.

Steve, your paper and our class discussion of forecasting clearly points out the issues and problems with using moving averages. Companies then compound the problem by assigning trendlines to a data presentation that is flawed from the beginning. The importance of tracking and trending data lies in the ability of the analysis to transpose the data to useful, relevant, timely, and accurate information, which can result in knowledge of the system or process being trended and in changes or improvements that can be made to that system or process. This class has been very useful in pointing out to us that we must use a system approach to bring about improvements and changes in our processes to achieve our ultimate goal of meeting and exceeding the customer's expectations. The use of control charts will provide us with a much better data representation, and hopefully will allow us to tranpose that information to improve our processes and to improve our bottom line. As I have said before, we must analyze each critical component of our system from supply of materials to input to our process to transformation of the materials into a product to output to our inventory to our customers. But ultimately, we must have a complete understanding of our customers needs and expectations, and monitor the data points in our system using statistical process control methodologies to achieve success.
 

Jim Wynne

Leader
Admin
Steve Prevette said:
We went over the "Forecasting" chapter, which makes heavy use of moving averages and regression. I would also like you to review the attached paper and provide any comments or real life sagas with such misuse of statistical principles.

See http://elsmar.com/Forums/showthread.php?t=9081 for the paper.

I’m a little late chiming in on this one, but I wanted to comment for two reasons. First, I think it’s an excellent article and highlights one of the most important components of what Deming referred to as “Profound Knowledge “: awareness of the power of basic statistical methods. :applause: As Steve points out in the article, some may quibble with the idea of applying normal-curve statistics to (potentially) non-normal distributions, but the basic information that might be derived is likely to be accurate enough to allow prudent decision making, which is what matters in most business cases. If students of business never learn anything else, they need to know how to use the rules of mathematical probability in decision making, and Steve’s article is a very articulate statement of the value of knowledge of basic statistical methods.

The other observation I have is a bit more arcane, but worth mentioning nonetheless. At the end of the piece Steve says that he used Excel functions to generate “random” numbers to form the data set. While in terms of the purpose of the illustration the use of the word “random” is accurate, in actuality it takes much more sophisticated apparatus than a personal computer to generate real random numbers. Most people are surprised to learn that computers are, in general, incapable of true random number generation, and the term pseudorandom came into being to describe what computers (and program algorithms) actually generate.
The way that Steve formed his Excel formula allowed the generation of numbers that were good enough for the demonstration, and as close as Excel can get to actual randomness. In most business cases it will be sufficient.

I won’t go into the bloody details here, but see the link below if you’re interested in learning more. Suffice it to say that if actual randomness is critical to your application, you need to be aware of the limitations of computers and software, and that the unexpected result of using pseudorandom numbers might be unreliable output.

Wikipedia: Pseudorandom number generators
 
T

tammye

figures & lies

Well stated classmates! It's sad but true that numbers (and resumes') can be skewed to make them look how the author intends. I work with a master spin doctor who can come up with ways to make numbers spin like you wouldn't believe. It's hard to imagine that census data, assessed valuation and salaries can be presented in less than factual ways, but it happens, as we all know. I think the important lesson to learn from this is to verify data and how the person came up with said data.
 
S

scrowner

forecasting

I thought the paper was very interesting. I have never taken the time to look at how the same data can look different-dependent on how it is graphed. I can see where the moving average would/could be used widely in all industries. This "tool" can make any data look good. I have just started to experience the use of control charts in the laboratory. We use these to watch instrument performance. When there is a sudden drop or rise in the response, I can actually go back to the logbook to see what may have caused this. I plan on comparing these to next years control charts to see if there is a seasonal trend to these environmental samples.
 
M

Mike Moran

Come on everyone, we see the same thing taking place evey day on the front of the USA Today newspaper. Take a look at whatever the data chart for the day is. There is always the fine print that tells a different story from what you thought you were looking at compared to the headline that goes with the chart.

Nice and very real comments/examples classmates. I like the resume one the best and the question of what does "experience" really mean when it is listed...

Steve,

Great paper. What a scary sense of reality....Guys like you could make or break a business with this stuff. Knowing the data, how it is generated and what story it tells is critical before making a management decision.
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
Yes, as was pointed out the Excel random number generator is actually a pseudo-random number generator. Computer pseudo-random number generators usually require some sort of "seed" value, then all numbers following are actually deterministic, but appear random. The Excel rand() function has never been documented in the open press, so if you are running a big simulation (per the simulation chapter) you may want to purchase a better quality generator.
 
Status
Not open for further replies.
Top Bottom