Reflective Essay

IS 301- 01

Donald Simpson

December 13, 2017

Help Received:

Online Dictionary

Online Thesaurus

See Works Cited

D. Chase Simpson

Reflective Essay

Over the course of the Fall Semester 2017 IS 301 served as my first true exposure to statistical analysis in political science. The first quarter of the course was largely refresher information but highly important none the less. Readings on theory, causality and conceptualization were critical because they set the foundation for the research project that took place during the second half of the semester. Of these initial topics, I found the discussion of causality uniquely pertinent to our group project. As I mentioned in my blog post on September 21^st, Jon Elster argued that in social science research analysis must be accompanied by developed theory with a strong causal mechanism to have any legitimacy. (Elster 1989) The importance of a strong causal mechanism was apparent in the early days of our group project.  Our research sought to examine how the source of cadet income determined their spending habits on products such as alcohol, tobacco, and fast food. Through an examination of past research on the spending habits of adolescents we decided to test a causal relationship between money from ROTC stipends and spending on these products. Our theory stated that this source of income provides both a windfall feeling, that cadets will not link income with hours worked, and a high degree of freedom to its use as the money is deposited straight into a cadet’s checking account and does not come from a parent or guardian who may have a degree of control on its use. We theorized that when combined these two factors would result in a higher propensity to consume in the previously mentioned items.

As the course transitioned from theoretical concepts in political research to the practical applications of using SPSS to analyze data and find correlation I experienced very little frustration with completing the worksheets. As I stated in my blog post of October 13^th, I found the workbook to be extremely easy to follow with step by step instructions including illustrations. I found SPSS itself to be not overly sophisticated but extremely user friendly, logical and perfectly capable to fulfill the requirements of the course. The one critical weakness of this learning method was that I often felt that I was often simply going through the motions to complete the assigned problem without taking the time to read and understand why I would use certain functions in different situations. This would have profound effects on the challenges I faced during the group project.

The last half of the course centered primarily on the group project. The group project provided for an opportunity to practice the concepts learned during the worksheets and provided myself ample opportunity for trial and error practice in using SPSS effectively. As stated in numerous blog posts my group found a degree of difficulty in creating an appropriate survey to collect data on cadet spending habits. Our first concern was to get a high participation rate while also collecting meaningful data. While we felt that we could get more meaningful data if we allowed cadets to simply write in their own answers to the questions, as suggested by LTC Sanborn, we also felt that we would have a much higher participation rate if we were able to condense the survey down to less than ten predominately multiple choice questions. Ultimately we chose to go with the latter but this was not without its problems. Poor questions, such as “how much did you spend last week?”, proved to be less useful than expected because of non-average spending habits in the holiday season. Also, a lapse in the numbering scheme of our surveys resulted in many cadets simply skipping the first, or first few questions on many surveys. Seeing that the majority of cadets had not written in their class year was exceptionally frustrating as that variable was supposed to be our primary control during the analysis. This problem was quickly resolved, however, when we began to sort the surveys by stoop as they were collected to determine from which class each cadet belonged. After the data had been inputted into the spreadsheet and coded, we sat out to conduct analysis. Despite having easily completed all the worksheets and learning all of the necessary functions of SPSS which I would use to conduct my analysis, my group and I were at a complete loss as to what to do. After a few frustrating hours I felt that we should wait and seek guidance during class the following day. This is where I began to feel all the concepts of the course coming together. I had learned the different uses of SPSS from creating new variables to running correlations, but it was only as LTC Sanborn described what needed to be done in the context of my group model was I able to understand why certain functions are used, and when.

In conclusion, the structure of the course, in most instances, set me up for success in completing the group project. My feelings throughout the course shifted from being interested in and challenged by writings and quizzes on causality and conceptualization to a perhaps a bit of overconfidence in my knowledge of using SPSS by easily completing the assigned worksheets. This overconfidence quickly transitioned into frustration during the group project as the realities of research began to set in. Poor wording in some questions and a questionable numbering patter on the survey led to significate confusion in how to fill out the survey and could have impacted our final analysis. Through the frustration of the group project, however, I was able to strongly increase my understanding of using statistical analysis effectively. Concepts began to make sense only once they were part of n research application and separated from the step by step instruction of the text book.

Works Cited

Darling , Helen , Anthony Reeder , Rob Mcgee, and Shelia Williams . 2006. “Brief report: Disposable income, and spending on fast food, alcohol, cigarettes, and gambling by New Zealand secondary school students .” Journal of adolescence 837-843.

Elster, Jon. 1989. Nuts and Bolts for the Social Sciences. Canbridge: Cambridge University Press.

Pollock III, Philip H. 2016. An IBM SPSS Companion to Political Analysis . Los Angeles : Sage.

Data analysis began with a slow start on Sunday morning the day after we have finished coding our data set. After running a few frequency distributions and cross tabs which I remembered how to do from the lessons taught during the course, I began to realize that I would be unable to complete the data analysis from memory and would need a bit of direction to begin. Therefore, on Monday during class I asked LTC Sanborn for a few suggestions on conducted the analysis. He first suggested examining the data collected using a Cross tab to determine if there were significant differences among the classes in spending habits and the source of spending. What we saw was that the modal category for the 1st class cadets was the highest category. Meanwhile for the 2nd class cadets it was the median category and for the 3rd class and rats it was the 2nd lowest category. The data on the source of income also correlated strongly with class as two trends clearly emerged. First, as cadets advanced through VMI they were more likely to have an ROTC stipend as their primary source of income and less likely to receive the majority of their money from their parents. Also, summer employment logically peaks before 3rd class year as it is the summer after matriculation least likely to impacted by summer military training. Because of these distinct differences in both spending habits and source of income among the classes, it was advised by LTC Sanborn they we create new variables for each Class and Source of income. Summer jobs and on post jobs were collapsed into one category when the new variables were drafted. 12 new variables were created as there were four classes and three possible primary sources of income (ROTC, job, or parents/family) These were coded as dichotomous. For example variable “S1_ROTC_18” was a “yes no” variable that showed whether a cadet in the class of 2018 had listed a ROTC stipend as their primary source of income.

As the data was collected by Cadets Moffatt, Gruber, and Elliott I inputed the data from the surveys into an Excel spreadsheet. After all data had been inputed we had 321 completed surveys. The variables, however, had been inputed into the spreadsheet with the original letter answer choices still in the data set. For analysis to be conducted through SPSS, however, these letters would have to be coded into numbers. On Saturday morning I established a coding pattern as follows. Yes No questions would be coded as Yes=1 and No=2. Multiple choice responses were coded in relation to their position in the alphabet. For example, A=1, B=2 etc. The data points were divided roughly equally between the four team members and were coded as stated previously. This process was quick and without any real problems. There was some debate on how we should code a few written in responses but it was decided that these data points were so few in number that they would have very little impact on the overall data set and they were left to be coded as “other”.

The problems with our survey design became apparent as the first round of completed surveys were delivered to me by cadet Elliott on the first night of data collection. The most noticeable problem was that cadets were not writing in their Academic Class at the top of the survey. I assumed that because many cadets were completing the surveys quickly and not paying a lot of attention when they were filling them out they were skipping the first question that was not numbered. Not numbering the question that asked what academic class each cadet belonged to was an oversight that should have been corrected before the surveys were distributed. This problem was quickly mitigated, however, as cadets conducting the surveys were asked to divided the surveys into different stacks based on what stoop they collected the surveys from. A second problem was that cadets did not fully understand or read the directions regarding ranking their sources of income. While many cadets did comply fully with the directions, several simply circled the sources that applied to them. This however, proved to be a minor problem and had little impact on the data set.

After getting approval from LTC Sanborn to begin distributing our surveys and collect data we began to develop a plan to collect our data. First, we had two feasible options for our data collection. We could either ask for our survey to be sent to the corps through email or we could conduct in person surveys in barracks. Reflecting on our own past tendencies to simply ignore surveys sent to us through email, we felt that participation rates would be higher if we approached cadets in person to conduct our surveys. Our original thought was for our group to collect our data early in the week between the hours of 7:30pm and 8:30pm. We felt that this would be the time period in which most cadets would be in their rooms. This plan quickly fell apart, however, as personal time constraints forced us to adopt a new plan for data collection. Instead of having set times to collect data we printed off 400 survey packets and established quotas for each member of the group that would be collecting data. We established that Cadets Elliott, Moffat and Gruber would collect data from cadets while I would staple the packets together and input the collected data into an excel spreadsheet as data came in. The group members were each assigned a barracks in which to collect data in and instructed to begin with first stoop and then move up.

After discussion with LTC Sanborn my group decided to make several edits to our survey. First we decided to collapse the categories of Family Household Income, Monthly Income, Percentage of weekly spending on Alcohol, Tobacco, and Fast food. Second, we decided to add a few more questions to collect data for possible controls in our data analysis. These controls include, whether a surveyed cadet is a NCAA athlete, and if they are the first member of their family to attend college. Lastly, the final question regarding the source of income was edited so that a difference between a summer job and a part time job during the school year was added. In addition, the instructions for the last question so that surveyed cadets rank their sources of income from 1-5 with one being their largest source of income. Below is the final survey that was distributed to cadets in all four classes in barracks.

Academic Class:

1. Gender
2. NCAA (Y or N)
3. Family Household Income
   • $0-50,000
   • $50,000-100,000
   • $100,000+
4.  First Generation to attend college
5. Monthly Income at VMI
   • $0-100
   • $101-200
   • $201-300
   • $301-400
   • $401+
6. Spending last week
   • $0-10
   • $11-20
   • $21-30
   • $31-40
   • $41-50
   • $51+
7. Percentage of weekly spending spent on Alcohol, tobacco (to include vape), and fast-food
   • 0-25%
   • 26-50%
   • 51-75%
   • 76%-100%
8. Source of income: Rank the following in order 1- Primary source of income – 5 smallest source of income, write N/A for all sources not applicable 
   • ROTC stipend
   • Part-time Job on post
   • Parent/ Guardian or other family
   • Summer Job
   • Other: List below

Help Received:

Guidance from LTC Sanborn

Brief report: Disposable income, and spending on fast food, alcohol, cigarettes, and gambling by New Zealand secondary school students. Helen Darling, Anthony I. Reeder, Rob McGee, Sheila Williams 2006

D. Chase Simpson

Below I have proposed a sample survey that we may adopt to collect data from the corps of cadets.  Parts of this survey are based on the survey conducted of New Zealand secondary school students in

“Brief report: Disposable income, and spending on fast food, alcohol, cigarettes, and gambling by New Zealand secondary school students” (Darling et.al 2006)

Academic Class

1. Gender
2. Family Household Income
   1. $0-25,ooo
   2. $25,001-50,000
   3. $50,001-75,000
   4. $75,001-100,000
   5. $100,000+
3. Monthly Income
   1. $0-50
   2. $51-100
   3. $101-150
   4. $151-200
   5. $201-250
   6. $251-300
   7. $301-350
   8. $351-400
   9. $401+
4. Weekly Spending
   1. $0-10
   2. $11-20
   3. $21-30
   4. $31-40
   5. $41-50
   6. $50+
5. Percentage of weekly spending spent on Alcohol, tobacco(to include vape), and fast-food
   1. 0-10%
   2. 11%-20%
   3. 21%-30%
   4. 31%-40%
   5. 41%-50%
   6. 51%-60%
   7. 61%-70%
   8. 71%-80%
   9. 81%-90%
   10. 91%-100%
6. Source of income
   1. ROTC stipend
   2. Part-time Job
   3. Parent/Guardian or other Family
   4. Other: List below

Help Received:

Brief report: Disposable income, and spending on fast food, alcohol, cigarettes, and gambling by New Zealand secondary school students. Helen Darling, Anthony I. Reeder, Rob McGee, Sheila Williams 2006

Over the last few weeks my group has begun to form how we will conduct our data collection. My first thought was to conduct our survey through a google doc emailed out to the corps of cadets. This model has a few advantages. First, the online survey would be less intrusive to people. They could complete the survey on their own time and perhaps be more willing to complete the survey due to that. There is, however, a distinct disadvantage to this model. I personally remember the many times I have completely ignored these emails without going any further than the subject line. Therefore, I believe it is most advantageous to conduct the survey in person. I believe that approaching people in barracks with a paper survey, would result in a much higher participation rate than conducting the survey through email. There would have to be several key guiding principles to this survey. First, the survey must be concise, no more than ten questions that can be conducted fairly quickly to encourage participation. Second, we must be careful to not bias our data to a specific class in barracks. It will be easiest to get participation among our classmates in the class of 2018. This however would not provide a representative sample of the corps. Third, while avoiding the potential for bias based on academic class we must also make sure we get a representative population for sources of income. Therefore, we must seek out equally large samples of commissioning and non commissioning cadets.

On our first day using the SPSS software LTC Sanborn described the product as the Toyota Camry of Statistical Analysis. In most ways I have come to agree with that. The process to purchase and download the software was at times frustrating and long but not overly difficult, much like a trip to a car dealership. Once up and running the software isn’t flashy but surprisingly reliable and user friendly. The accompanying workbook, “An IBM SPSS companion to Political Analysis” by Phillip Pollock III provides uncomplicated, easy to follow instructions to complete basic but practical statistical analysis. I have found the work book quite helpful when I have been stuck on problems but for the most part the I have found the software itself efficiently set up to the point where I could, if needed, search through the menu and figure out how to run my tests independent of the workbook.

With the reality of difficulties in defining, let alone testing, the concepts in our original question setting in, our group decided to change course almost entirely and propose a new question. During one of our initial brain storming sessions Cadet Elliott offered an alternative question, “what is the relationship between Cadet disposable income levels and the amount of alcohol that they consume.” This struck my attention immediately. It appeared to be interesting, definable, and testable in the context of this class. After my initial meeting on the subject with LTC Sanborn, however, I began to agree that it was perhaps too narrow of a concept. We began to discuss the idea that perhaps whether a cadet is commissioning or not impacts if they spend more money on alcohol. This research question, while more explanatory still had several limiting factors, first it was only dichotomous in its source of money, commissioning or non commissioning. This is severely limiting as there are numerous ways for non-commissioning cadets to gain disposable income, (i.e. Part time job, allowance from parents, etc.) Second, it was perhaps testing something different than what its intention was. If tested as originally stated it would simply answer whether or not having more money allows a cadet to purchase more alcohol an obvious correlation already exists without a test, instead there need to be a better test perhaps as a percentage of income. Finally, the testing of just alcohol was severely limiting, there are many people who don’t consume alcohol for reasons other than lack of funds. In addition, the legal drinking age of 21 limits the testable group to predominantly first class cadets.