Mathematics & Statisticshttp://hdl.handle.net/10211.3/1224902018-06-25T05:46:01Z2018-06-25T05:46:01ZDiscrete-Time Bayesian Survival Analysis: The Impact of Stress Management CoursesDelgado, Rafaelhttp://hdl.handle.net/10211.3/2012522018-03-21T21:35:41Z2018-03-21T00:00:00ZDiscrete-Time Bayesian Survival Analysis: The Impact of Stress Management Courses
Delgado, Rafael
Students fail to complete undergraduate degrees from universities for a variety of reasons. Although many factors exist that impact a student’s decision to leave university, common factors include stress, depression, burnout, anxiety, work, and financial difficulties. Some students may decide to stop out from university, putting their education on hold with the intention of returning and completing a degree. While universities offer health and counseling services to address students’ physical and mental well-being, many students do not seek help from counseling services when they need it. At Cal Poly Pomona, a growing number of undergraduate students have enrolled in stress management courses offered through the Department of Kinesiology and Health Promotion. In this thesis, we will use Bayesian survival analysis to assess the impact that a stress management course has on graduation rates by comparing students who enrolled in the stress management course to those who did not enroll in the course. Results show that students who enroll in the course attain graduation rates higher than those who do not take the course, particularly students who enroll during the early quarters of junior year. Additionally, students who enroll in the stress management course the first quarter of their junior year attain three- and four-year graduation rates higher than students who never enroll in the course when taking cumulative grade point average into account. It is important to note that the association between taking stress management courses and graduation rates should not be interpreted as a causal relationship between the two events.
2018-03-21T00:00:00ZBig Data in the Big AppleCostello, Dylanhttp://hdl.handle.net/10211.3/2003792018-02-24T00:31:39Z2018-02-23T00:00:00ZBig Data in the Big Apple
Costello, Dylan
Today we have a greater supply of large data than ever before, and with this increase in large data has come a demand for analysis of large data. As a data set increases in size it becomes increasingly difﬁcult to perform statistical analysis. Data needs to be properly cleaned, and code must be written to make computer computation efﬁcient. Even when this is done, computation may still be too heavy for a standard personal computer. In this event, computation can be outsourced to a super computer or cluster of computers through a server. In this thesis, we analyze a large data set of taxi rides from the Yellow Cab company. This is a public data set distributed by the New York City Government and is roughly 207 gigabytes in size. There are numerous ways this can be done, many of which will be covered in this thesis. The factors in deciding the method of analysis include budget, timeline, and hardware. Thosewithalargebudgethavetheadvatageofusingexpensivehardwarethat yieldfastresults. Forthisgroupofpeople,cloudcomputingcanbeusedtorentpowerful hardware. However, cloud computing is not reserved for those with deep pockets. The idealmethodmayinvolveexpensivehardware, butalternativemethodscan beemployed on relatively inexpensive hardware rented from a cloud computing company. For those without the necessary knowledge to utilize cloud computing, some analysis can be done on a personal computer. Although, this is not the ideal method when working with large data. A portion of the analysis will be done on a personal computer to understand what can be done without cloud computing. This includes cleaning the data, analyzing some trends,andbuildingaknnalgorithm. Theremainderwillbedoneusingcloudcomputing,which gives an idea of the capabilites of cloud computing versus the capabilites of a personal computer. With cloud computing we will do some general data exploration, build many predictive models, and then compare their level of accuracy. The models include linear regression, regression trees, and random forest.
2018-02-23T00:00:00ZAssessing Blinding in Randomized Clinical TrialsWaite, Jesse C.http://hdl.handle.net/10211.3/1966992017-10-04T19:37:40Z2017-10-04T00:00:00ZAssessing Blinding in Randomized Clinical Trials
Waite, Jesse C.
In the realm of randomized clinical trials, protocols intended to protect the knowledge pertaining to which treatment assignment each participant actually receives are usually employed. These protocols promote what is known as \textit{blinding}. When these protocols are meant to obscure the assignment from the clinicians as well as the participant, this is known as a double blind study. It is widely held that successfully employing protocols to insure blinding will help to insure the results of the study are not subject to bias. This thesis will discuss some of the methods commonly included in the protocols regarding blinding and the assessment of its success as it pertains to randomized clinical trials. Three methods which have been and could be used to assess the success of blinding protocols will be analyzed. A simulation study comparing the three methods for assessing blinding using R will show the differences between these methods, and their strengths and weaknesses will be discussed. Finally the development and employment of a method for determining when unblinding occurs because appropriate protocols are not enacted or followed is discussed.
2017-10-04T00:00:00ZTridiagonal Stochastic MatricesNguyen, Uyenhttp://hdl.handle.net/10211.3/1963472017-09-26T19:56:02Z2017-09-25T00:00:00ZTridiagonal Stochastic Matrices
Nguyen, Uyen
A birth-death chain with one-step transition probability matrix P often has a dual birth-death chain with one-step transition probability matrix P^*. The same holds for birth-death processes. From Professor Kouachi's work, we are able to determine the eigenvalues of suitable matrices P and P^*. We describe the exact diagonalization of P and P^* in Chapter 1. Chapter 2 summarizes Professor Kouachi's work in determining the exact formulas for eigenvalues and eigenvectors of certain tridiagonal matrices having arbitrary large dimension.
In Chapter 3, we apply Professor Kouachi's results to diagonalize a certain class of birth-death chains and processes. We obtain exact expressions for P^n, (P^*)^n and P(t). Generalizations of our results to non-tridiagonal stochastic matrices are presented in Chapter 4. Final conclusions and plans for future work are given in Chapter 5. Computer programs written by collaborators by Luis Cervantes, Mark Dela and Dave Luk to calculate P^n, (P^*)^n and P(t) are gratefully acknowledged and used in this thesis to obtain results in higher dimensions.
2017-09-25T00:00:00Z