The project goal was to measure the age of the star cluster NGC 2516 through data of the eclipsing binary system V392 Carinae. Photometric data by the Transiting Exoplanet Space Telescope (TESS) and spectroscopic data by European Southern Observatory (ESO) were used to measure precise mass and radius measurements of the stars. During the process of modeling our two star system through the Eclipsing Light Curve program (ELC, fortran code by Orosz and Hauschildt 2000), ​we discovered a third star in the system. Python codes were developed to process and analyze data from 1978-1991, 1997, 2004, 2006, 2009, and 2019 using Numpy, Matplotlib, Astropy, Pandas, lightkurve, among others libraries.

The Hubble constant is a fundamental parameter that sets the scale and age of the universe. Time series images of 9 supernova were taken over 6 weeks using the Mount Laguna Observatory. Data was corrected for instrumental, atmospheric and other forms of contamination. Resulting light curves were chi square minimized to fit to known supernova light curve models. The measured supernova distances and receding velocities were least square fitted to get Hubble’s constant. Python codes were developed to process and analyze data using Numpy, Matplotlib, Astropy, and Pandas.

5GB of Gaia (space satellite) data were used to predict the radial position of a star from the center of the Milky Way. Calculations were handled using pyspark in AWS EC2. A Decision Tree Classifier was used from the Pyspark ML library with 2 predictor variables of surface temperature of the star (Teff) and the brightness measurement (Gmag ). We found that a wider population of stars are seen in the spiral arms of the galaxy, especially younger bluer stars.

The purpose of the study was to understand the percentage of subsidized meals in California schools using the Academic Index for California Schools data set. I developed R codes to do 10,000 sample simulations based on the observed data using three sampling methods: simple random sampling, proportional stratified sampling and one stage cluster sampling. I computed the mean percentage and standard error of subsidized meals using each sampling method. I found a similar mean percentage of subsidized meals for the 3 different sampling methods. However, the standard error is wider in one stage cluster sampling because the dataset is biased towards more elementary schools. This resulted in a bivariate distribution for the clustering method.

The goal of the project was to build a model that will facilitate bee classifications for an entomologist friend. 504 images of isolated bee wings of 150 by 60 pixels in three channels were used for classification. Three ML techniques were used to categorize three genuses based on their wing venation patterns. The SVM rbf kernel gave 44% accuracy, the logistic regression model achieved 52% accuracy and the CNN using TensorFlow gave 60% accuracy. Larger sample of images were needed for better results.

During the 10 week program I mentored 2 high school students in a research project related to my thesis project (above). Along the way they learned to develop python code of their own. Both students went on to be semifinalists in California’s Siemens competition.

Develop python code to measure galaxy rotation and dispersion profiles using Keck DEIMOS spectroscopy data. Multi gaussian models with varying amplitudes and widths were used to find velocities of the galaxies. This allowed for the categorization of regular or chaotic systems. We found that chaotic systems are prominent in the past & uniform ones are common in the present.