9-2 Final Project

David Velez

Southern New Hampshire University

QSO 510- Quantitative Analysis

Professor Dereshiwsky

January , 2025

I: Introduction

A. For this scenario, I will be analyzing the response variable (GIFTAMNT), which is the amount of money donated by donors, and focus on male veterans (MALEVETS). The variable of male veterans is based on the donor’s zip code. As seen in the three histograms below, the frequency with the highest amount of money donated by GIFTAMNT ranges between 0 and 25 with the frequency of histogram one being 1,732 and histogram two being 1,323. The problem identified in these histograms is that the amount of money being donated drops drastically after 25. When looking at histogram three for male veterans, the highest frequency is 825 and that ranges between 30-35.

A screenshot of a graph

II: Analysis Plan

As it states in the textbook, “to generate donations, the PVA sends out greeting cards and mailing address labels periodically with their request for donations.” (Sharpe, 2019). With this approach, donations are being sent but can there potentially be other methods so more donations can be sent? The best way to answer “yes” to this question would be taking a marketing approach and investigate a variety of methods that are affordable and most importantly realistic for both the institution responsible for receiving and advertising the donations and the donors.

Donations are being received for male veterans through the current method which involves greeting cards being sent out by mail but is there a possibility of receiving a lump sum donation to improve the amount received beyond 50?

Since these donations are based on the donor’s zip code, one method to help with more donations can be partnering with local channels and advertising commercials on tv for the donors. Another method that can be used is sponsoring ads on social media platforms like Instagram, Facebook, and TikTok. Lump sum donations can play a major role when it comes to social media platforms like the ones stated previously because groups can be created and many people can participate. With this method, a Facebook group would be a great example because someone can oversee running the page/group and Facebook users can then donate whatever amount they desire, just like they do with the greeting cards they receive in the mail. After a certain amount of time, for example the last Friday of every month, that specific person who is overseeing the Facebook group can then gather all the donation data and make a lump sum payment to improve the statistics. Mailing these cards did cost PVA over “$40 million in postage, administrative, and gift expenses.” (Sharpe, 2019) and there were people who never even responded. Something that can be done is instead of spending over $40 million in sending out greeting cards by mail is cutting that by half and use the other $20 million for things like tv commercials, social media ads, and organizing donation groups.

III. Statistical Methods

A. Selected Statistical Methods and Assumptions

To analyze the relationship between the company’s fundraising activity and veteran engagement outcomes, multiple linear regression will be used as the primary statistical method. In this model, MALEVET serves as the dependent variable, representing the number of male veterans engaged by the organization, while GIFTAMNT is the independent variable, representing the monetary amount of gifts received. Multiple linear regression is appropriate when the objective is to evaluate whether changes in one or more predictor variables explain variation in a continuous outcome variable (Ahmed et al., 2024). Multiple linear regression was selected over descriptive or correlation-only methods because it allows for prediction, hypothesis testing, and quantification of effect size, all of which are critical for managerial decision-making. Even when a relationship is weak or statistically insignificant, regression analysis provides insight into whether operational assumptions about drivers of outcomes are supported by data.

One of the underlying assumptions of the multiple regression is that the relationship between the gift amount and veteran engagement is linear. In addition, the model assumes that there is minimal multicollinearity since there is only one explanatory variable for the response variable (Hilbe, 2025). The model also assumes that the observations are independent implying that each data point originates from a distinct donor or transaction. The Multiple regression also assumes that variance of errors across predicted values is constant and that the residuals are normally distributed. According to Lee et al. (2022), normal distribution is assumed for large sample sizes, and in this case, n=3648.

B. Justification

Regression analysis supports evidence-based decision-making by allowing managers to test whether financial inputs meaningfully influence organizational outcomes. In this case, the regression results indicate the model MALEVET=31.64−0.022(GIFTAMNT) as shown in figure 1 below.

Figure 1

Regression analysis results

While the intercept is statistically significant p < 0.0001, the coefficient for GIFTAMNT is not statistically significant p = 0.1456. The model’s coefficient of determination (R^2) value of 0.0006 indicates that gift amount explains 0.061% of the variance in male veteran engagement. The overall F-test confirms that the model does not significantly improve prediction over the mean response. The key actionable insight from this analysis is that financial contributions alone do not predict engagement levels. As such, decision makers can stop overinvesting in fundraising strategies as they are unlikely to yield operational improvements.

IV. Data-Driven Decisions

A. Decision-Making Process Using Statistical Analysis

The decision-making process begins by defining whether financial donations influence male veteran engagement. Relevant variables were operationalized, and regression analysis was used to test statistical significance (Charles et al., 2023). The findings indicate that gift amount does not have a strong predictive power which implies that increasing donations alone is may not translate to improved engagement outcomes. As such, a revised decision framework should incorporate additional variables and qualitative insights.

B. Use of Data Mining and Metadata

Regression analysis serves as an initial supervised data mining technique. Additional mining methods, such as clustering or classification, may help identify non-financial drivers of engagement. According to Han et al. (2022), metadata is essential for documenting variable definitions, data sources, and measurement periods, ensuring accurate interpretation and consistent application of analytical results across the organization.

C. Structured vs. Unstructured Problem

The problem is primarily structured, involving defined variables and quantitative analysis (Song et al., 2022). However, engagement outcomes may also be influenced by semi-structured factors, such as motivation or outreach quality, which are not fully captured by numerical data. Managers should be prudent enough to supplement statistical findings with qualitative judgment.

D. Assessment of Variable Potential

While MALEVET captures engagement outcomes, it is influenced by factors beyond financial contributions. GIFTAMNT represents monetary input but lacks contextual depth. The model shows a low explanatory power, which implies that there are additional variables, such as outreach methods, demographics, or service accessibility, that can better explain engagement levels (Ahmed et al., 2024). This underscores the importance of effective variable selection

V: Recommend Operational Improvements

A. Data-Driven Calculation

B. Summarize Results

C. Solution

References

Ahmed, R. R., Streimikiene, D., Streimikis, J., & Siksnelyte-Butkiene, I. (2024). A comparative analysis of multivariate approaches for data analysis in management sciences.  E+M Ekonomie a Management.,  27(1), 192-210. https://doi.org/10.15240/tul/001/2024-5-001

Charles, V., Garg, P., Gupta, N., & Agarwal, M. (2023). Data Analytics and Business Intelligence.  Data Analytics and Business Intelligence.

Han, J., Pei, J., & Tong, H. (2022).  Data mining: concepts and techniques. Morgan kaufmann.

Hilbe, J. M. (2025). Generalized linear models. In  International Encyclopedia of Statistical Science (pp. 1035-1042). Berlin, Heidelberg: Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-662-69359-9_250

Lee, C. S., Cheang, P. Y. S., & Moslehpour, M. (2022). Predictive analytics in business analytics: decision tree.  Advances in Decision Sciences,  26(1), 1-29. https://doi.org/10.47654/v26y2022i1p1-30

Sharpe, N. R., De Veaux, R. D., & Velleman, P. F. (2019). Business statistics (4th ed.). Pearson.

Song, B., Zhang, Z., Qin, Y., Liu, X., & Hu, H. (2022). Quantitative analysis of freight train derailment severity with structured and unstructured data.  Reliability Engineering & System Safety,  224, 108563. https://doi.org/10.1016/j.ress.2022.108563