About the Author
Zhihao (Howard) Yao
A skilled data scientist with over 15 years of expertise in data management, analytics, and visualization, bringing a unique combination of government experience as a mathematical statistician at the FDA and industry experience as a Senior Principal Data Scientist at a leading medical device company.
Proven ability to deliver impactful insights and drive data-driven decisions across diverse environments. This diverse expertise spans across regulatory settings and industry-driven innovation, consistently applying advanced analytics to inform critical decision-making.
Proven ability to deliver impactful insights and drive data-driven decisions across diverse environments. This diverse expertise spans across regulatory settings and industry-driven innovation, consistently applying advanced analytics to inform critical decision-making.
Demo Projects Portfolio
Statistics | Likelihood Ratio-Based Tests for Drug/Device Data Monitoring
Likelihood Ratio-Based Test (LRT) Method
introduces basic LRT method and applications of LRT methods to drug/device data in post-market safety surveillance.
introduces basic LRT method and applications of LRT methods to drug/device data in post-market safety surveillance.
LRT Application Part 1: Data Wrangling
The goal is to create an adverse events data frame for further LRT analysis after querying and wrangling the JSON files from FAERS database.
The goal is to create an adverse events data frame for further LRT analysis after querying and wrangling the JSON files from FAERS database.
LRT Application Part 2: Signal Detection and Visualization
LRT method, via an extensive simulation study, retains good power and sensitivity for identifying signals.
LRT method, via an extensive simulation study, retains good power and sensitivity for identifying signals.
LRT Signal Analysis Application
LRT Signal Analysis Application with dynamic parameters, such as name of drug, date range, and max number of events.
LRT Signal Analysis Application with dynamic parameters, such as name of drug, date range, and max number of events.
Statistics | Propensity Score
Leveraging External Evidence for Augmenting clinical Study
Propensity score-integrated power prior and PS-integrated composite likelihood approaches for leveraging real-world data in clinical studies.
Propensity score-integrated power prior and PS-integrated composite likelihood approaches for leveraging real-world data in clinical studies.
Propensity Score R Package MatchIt Introduction
MatchIt provides a simple and straightforward interface to various methods of matching for covariate balance in observational studies.
MatchIt provides a simple and straightforward interface to various methods of matching for covariate balance in observational studies.
Machine Learning
ML | 1 Introduction, Math, and Statistics
- Frequentist vs Bayesian
- Maximum likelihood Estimation (MLE)
- Statistics: Basic Concepts
ML | 2 Linear Regression
- Least squares estimation
- Regularization
- Maximum a posteriori probability
ML | 3.1 Linear Classification: Perceptron & LDA
- Perceptron
- Linear Discriminant Analysis
- Loss function
- Gradient descent
ML | 3.2 Linear Classification: Logistic Regression & NB & GDA
- Logistic Regression
- Naive Bayesian
- Gaussian Discriminant Analysis
ML | 4 Dimensionality Reduction
- Centering Matrix
- Principal component analysis (PCA)
- PCA Loss function
ML | 5 Support Vector Machine
- Max margin
- Optimization
- KKT conditions
- Soft Margin SVM
ML | 6 Decision Tree
- Information Theory
- Information Gain
- GINI Index
- Classification and Regression Tree
ML | 7 Ensemble Learning
- Bagging
- Boosting
- AdaBoost
- Gradient Boosting
ML | 8 XGBoost
- Elements of Tree Learning
- Taylor Expansion Approximation of Loss function
- Structure Score
- Greedy Learning of the Tree
ML | Case Study - Unleash the power of Python plus JavaScript
- Integration of Python and JavaScript,
- Descriptive Statistics with dynamic visualization
- Data preprocessing
- Model Training
ML | Object Detection Part 1: Custom YOLO V4 Model Training
Train a custom face mask detection model by using an AI computer vision system YOLO v4.
Train a custom face mask detection model by using an AI computer vision system YOLO v4.
ML | Object Detection Part 2: Instant Face Mask Detection with Webcam
Create an instant face mask detection application by using the webcam on desktop/laptop.
Create an instant face mask detection application by using the webcam on desktop/laptop.
Visualization
Visualization Selection Diagram
Visualize the four key pillars of data visualization: distribution, relationship, composition, and comparison, along with the specific charts used in each category.
Visualize the four key pillars of data visualization: distribution, relationship, composition, and comparison, along with the specific charts used in each category.
Covid-19 Global Dataset Visualization
A ridgeline plot offers greater precision within a limited vertical space by stacking distributions, though it comes with the disadvantage of potential occlusion. This issue can be effectively addressed by adding dynamic click events, allowing users to display or hide sections of the plot as needed, providing greater clarity and interactivity.
A ridgeline plot offers greater precision within a limited vertical space by stacking distributions, though it comes with the disadvantage of potential occlusion. This issue can be effectively addressed by adding dynamic click events, allowing users to display or hide sections of the plot as needed, providing greater clarity and interactivity.
Sequence Alignment Dashboard
Protein sequences alignment dynamic dashboard. Mini-map can be used as a brush selector and concensus is made of the model of Amino Acids of proteins.
Protein sequences alignment dynamic dashboard. Mini-map can be used as a brush selector and concensus is made of the model of Amino Acids of proteins.
TOP 30 Medical Device Companies
Medical Product Outsourcing (MPO) magazine published "The 2021 MPO Top 30 Medical Device Companies Report." This visualization provides an in-depth look at key metrics for these companies, including total revenue, profit changes, revenue per capita, and more, offering a comprehensive overview of their performance.
Medical Product Outsourcing (MPO) magazine published "The 2021 MPO Top 30 Medical Device Companies Report." This visualization provides an in-depth look at key metrics for these companies, including total revenue, profit changes, revenue per capita, and more, offering a comprehensive overview of their performance.