Statistics in Life
It's just everywhere!
When asked for some advice on pursuing a doctorate in statistics, my advisor told me this, "To have a Ph.D. in statistics is to have a backyard access to every field of science." Statistics is an irreplaceable tool and the language of science itself, enabling inductive reasoning in science and allowing researchers to uncover the true law under the noise.
Below are examples of statistics used in social science, network engineering, and political science. There are boundless examples of interdisciplinary collaboration, but these are the ones that I've had the pleasure of working on so far.
Inadequacies in Self-Evaluation
Accurate self-assessment regarding one's competence in completing a task is a critical life skill that is often not given the recognition it deserves. With it, the challenges one can take can be more manageable, the negative side effects of hubris can be counteracted, and society can run more efficiently. For this study, individuals were provided with a set of 25 questions and were asked to rate how confident they are at being able to solve the problem. Afterward, the same individuals were asked to solve said problems to measure how accurate their self-assessments were.
Analysis of the dataset provided by Prof. Ed Nuhfer uncovered interesting results. Statistical tests support the existence of the Dunning-Kruger Effect, which states that relatively unskilled people suffer illusory superiority and mistakenly assess their abilities to be much higher than they are. Additional discoveries suggest that women tend to underestimate themselves more compared to men but are more accurate in their self-assessments overall. Students and professors who majored in or dealt with science-related fields tended to be more accurate than students from other areas of study. The study also shows that by participating in the study, individuals can improve their self-assessments even without the evaluations of their self-assessments. The variables including race/ethnicity, region, type of institution, and other specific areas of study remain to be explored in future studies.
Capturing the Loss: Internet Package Transfer
The objective of the analysis was to predict the future packet loss rate and to identify any variables that might impact the packet loss rate. The dataset includes a total of 147 observations with variables Run Number, First Flow Start Time, Flow Start Time, Duration, RTT, User Send Rate, Achieved Rate, Flow Loss Rate, and Background Traffic Loss Rate. Logistic regressions, as well as Poisson regression, were performed to identify the model of best fit. The analysis returned a model of best fit within individual runs as well as a suggestion to isolated the variables to more effectively determine their impacts.
What's the Size of Your Cabinet? Cabinet size and the chance of Exile
The primary objective of this statistical analysis was to determine the presence of a significant association between a cabinet size available to any given African leader and their possibility of forced exile. The dataset contained observations from 43 African countries and their leaders along with years of continuous leadership, available cabinet size for the given year, presence of civil war, polity score, presence of oil as over a third of GDP, presence of mineral as over a third of GDP, presence of ethnic fractionalization, lag GDP, and log of population. The dataset was analyzed primarily through exploratory data analysis and linear regression methods, which suggested that there indeed is a significant relationship between available cabinet size and the likelihood of forceful exile. For countries that do not depend on oil or mineral as a major portion of their GDP, data suggests that for every additional cabinet position available, the likelihood of a leader's exile decreases.