Evaluated the efficacy of a novel drug for multiple myeloma (ND314), compared to the standard of care(SOC).
Implemented a two-stage randomized superiority trial to assess the efficacy of ND314 by conducting stochastic simulations to validate the sufficiency of type I error and power for the proposed sample size and study design.
Proposed and justified a trial design including permuted block randomization, double-blinding, and the option for an early termination in case of overwhelming evidence favoring one treatment, while suggesting possible modifications.
A two-stage clinical trial for ND314 on Multiple Myeloma
Explored the efficacy of Ribociclib as a potential treatment for glioblastoma by comparing average phospho-Rb levels of mice between treatment and control groups.
Employed a nested linear mixed model to analyze repeated measurements of phospho-Rb levels, as well as a linear regression model to validate phospho-Rb as a surrogate for survival.
Found that the average phospho-Rb level of mice in treatment groups involving Ribociclib is significantly lower than that of mice in the control group, providing evidence in favor of Ribociclib, through survival and phospho-Rb are not significantly associated.
R packages used: dplyr, tidyr, reshape2, lme4, ggplot2, nlme, geepack, lme4, lmerTest
Evaluation of efficacy of CDK4/6 inhibitor for treatment glioblastoma using patient-derived xenografts
Analyzed data from the National Longitudinal Survey of Youth through linear mixed models to examine the difference in time trends of academic success across different geographic environments in the US.
Found that regional disparities in GPA trends are significant, while differences between urban and rural classifications are not.
R packages used: EnvStats, ggplot2, knitr, nlme, dplyr, reshape2
Time trends of academic performance by geographical environment within the United States
Designed a study on the effects of the interventions, HELP and Family-HELP, for post-operative delirium, employing the stepped-wedge design and analyzing the data using linear mixed effects models.
Implemented a stepped-wedge design comparing the effects of two interventions for post-operative delirium: HELP and Family-HELP involving clusters of subjects receiving the interventions at different times, ensuring everyone receives the intervention.
Utilized the linear mixed effects model assuming exchangeable correlation and equal cluster sizes, and planned GEE modeling for when cluster sizes are not equal.
Stepped-wedge design for prevention, early identification, and treatment of delirium in older adults
Course projects
Data analysis projects using R, Python, and Tableau. Other projects through Figma and Flutter.
Interested in differences between United States and South Korea.
Veganism and LGBTQ+ were two key differences, and gathered data regarding those to topics.
Creating 'progressivism index', concluded veganism and LGBT community has strong relatiionship.
Progressivism index by state upon veganism trend, LGBT community, feminism index score, etc.
Doordash data from Kaggle, and analyze data referring to Ben Roshan's analysis
Through demographic analyses, time factor analyses, univarite analysis, can conclude that time of delivery and rating of restaurant plays a major role for customer satisfaction.
Food delivery service, Doordash Exploratory data anlaysis
Spotify track dataset with 21 variables, analyzed following Hossein Faridnasr's instruction from Kaggle.
Split dataset into a training and test dataset.
Comparing models, and XGBoost model performs better.
Regression analysis and popularity prediction with XGBoost
New York State Airbnb price by neighbourhood; Manhattan, Brooklyn, Queens, Bronx, and Staten island.
TreeMap and bar graph using Tableau.
Total and average value of Airbnb by neighbourhood
Side projects
Didim is one of the biggest food and beverage companies at South Korea.
Gathered 200 people’s user experience survey by questionnaire our team made about ‘Mapo Galmaegi’ , K-bbq franchise restaurant, and proposed new marketing strategy analyzing survey.
Executed proposed marketing strategy to 3 main branches and increased their profits.
Ranked second.
Didim marketing strategy competition
CU is a convenience store owned by BGF Retails, and BGF Retails is one of the biggest retail companies at South Korea.
Surveyed 250 user experiences of CU’s mobile application , second most profitable convenience store chain in South Korea and compared it to user experiences of GS25’s mobile application, the most profitable convenience store chain.
Devised sample arcade game using html which actually increased monthly active user of CU’s application.
Ranked third.
CU membership application competition
BnK Lab is a social dating service at South Korea like Tinder.
Designed and filmed two promotion videos and both were in use as advertisement on Youtube and Facebook.
Winner of the competition.