Provide me half page response to each discussion post. For the Discussion #1 and Discussion #2 tell me what you liked and what you disliked about the post and provide any alternatives of each post. For Discussion #3 answer the questions that are presented in detail.
Discussion #1:
To reiterate, my workplace is a College where I work in the admissions department. My daily objective is to enroll students into the school and help them through the enrollment process. Part of our daily activities is to forecast our expected enrollments each day, week and quarter. We in admissions can base our forecasts off of various components to determine accurate numbers.
I do think that using decomposition would be appropriate in my field. As I mentioned previously, there are many factors or components that can go into our forecasts. These factors can be based on season, the economic state, government funding/regulations and marketing efforts. Each of these components can increase or decrease demand for higher education by itself. So, if we were to focus on each component individually, we would be able to assess part of our forecast based upon each component. For example, if we look at the various seasons of the year. Our fall term always has a naturally higher number of students inquiring about school. Our summer term has the least. As a college we can better make our forecasts on that component alone, which will better prepare our college for a specific number of enrollments.
One challenge by using decomposition is that we may forget to take into consideration other factors. If we make our forecast off of just one of the components, say for example seasonal, we may miss another big component. Let's say our fall term is coming up and we forecast a high number of enrollments. However, the federal government has just cut student funding, which could result in higher expenses for students. If this is the case, then we just forecasted high enrollments because of the season, when in reality the forecast should be much lower due to the affect government funding will have on the enrollment numbers.
For longer term forecasting, I do think that decomposition would be applicable. This is the type of forecasting that would be done much higher up in the College than on the ground level of admissions. Directors and managers could plan long term for the college by studying certain components. Trend, government regulations and economic growth can be these components. If the state of the economy is getting better and is projected to get much better in the next 5 years, than as a college we can expect a steady decline in enrollments as the economy grows stronger. I also do think that the seasonally adjusted data would work. In my field, we can definitely see how the seasons affect student enrollments, so this would factor in nicely.
Discussion #2:
To review, my employment is at a university where my primary function is to work within the admissions department. My previous posts have been regarding the enrollment numbers for a particular quarter. Even though my organization has a tough time separating the target/goal definition with the forecast definition; the method of forecasting we use is time series.
The 2 largest challenges we face in applying a decomposition model is understanding the components and how it will affect our forecast. We typically see the largest enrollment numbers in the fall due to demand market. This makes our time series seasonal and it is difficult to determine if this seasonal fluctuation is due to outside sources; weather, timing, or geographical location. The second area of determination is the cyclical component. We notice increase in ease of enrollments when the unemployment rate is climbing and decrease ease of enrollments when the unemployment rate is falling. The trend and irregular component is equally challenging to determine. Is there a population change in our area? Are there more high school graduates this year than last year? Will the department of education (DOE) change federal loan rates, eligibility, and amount? These are all questions that are very hard to determine. If we are unable to answer these questions we will be unable to determine how these base components will be affecting our end forecast number.
Using any type of forecasting method will be better than not using one at all. The decomposition method is what we have used in the past and we continue to use this type of method (whether we are all aware of it or not). When we are able to answer the unemployment, DOE, population, and marketing efforts we are able to produce a fairly accurate long term model. We look at each individual component and how it will relate to the end forecast. I find that this decomposition method is great for identifying trend, seasonal, and cyclical factors, but falls short on the end forecast.
All data is useful, it is how we turn it into information is the hard part. Upon review of the readings and PowerPoint I find that our answers to some of the above questions may not always fit in the decomposition method. The multiplicative components model is closely related as described above. As time increases there is more variability to our observations; unemployment rate, DOE decisions, population in target areas, and marketing. Furthermore, we clearly have a quarterly series and a trend that defines a seasonal pattern. The Multiplicative model would work well with this component setup.
Discussion #3: Read and Answer the following questions for this one:
Let's talk a little bit about accuracy. When you fit a decomposition model to data using Minitab, you are provided with accuracy measures (specifically, MAPE, MAD, and MSE). Explain what these measure. Are they the same as forecast accuracy measures? Why or why not.