In this part 2 project, we add part 2 to part 1 separated by equal signs and 3 stars after the QC, filtering, normalizing, getting high variability genes, clustering with KNN and UMAP and TSNE, then get fold change values in part 2 for top 20 plus top 10 in Seurat's algorithm and add those FC values as well, test the significance in predicting the class type of GC or PM for Gastric Carcinoma or Peritoneal Metastasis, and scored 100% accuracy on both sets of genes, but 100% accuracy on the training and testing hold out validation set for the top 20 fold change values after omitting 0.000000 values after removing NAs and Infinites. Then added them to our pathologies database. Links in document also to the Tableua dashboard on FCs for each pathology we analyzed so far just by FCs related to EBV, and not but close, Fibromyalgia, Lyme disease, EBV infection, mononucleosis (only one in miRNA and no genes the same in other sets), multiple sclerosis, Hodgkin's Lymphoma, Natural Killer T Cell Lymphoma, Gastric Carcinoma, and HIV infected Hodgkin's with EBV, and uterine fibroids. We will see after gathering more data how well a model can be tuned with these top genes of fold change values to predict pathologies or show their similarities across pathologies by gene affects from disease.

The dependent variable is the mean learning gain (9th grade score minus 5th grade score) for that race group in that municipality-cohort. The analysis runs OLS with progressively richer controls, separately for 6 subgroups (3 races × 2 SES halves), then computes state value-added from Model 3.

Laporan ini menyajikan studi perbandingan antara lima algoritma clustering utama: K-Means, K-Medians, DBSCAN, Mean Shift, dan Fuzzy C-Means dengan menggunakan dataset Diagnosis Kanker Payudara (WDBC Dataset). Analisis difokuskan pada kemampuan masing-masing algoritma dalam mengelompokkan karakteristik morfologi sel menjadi kategori jinak (benign) dan ganas (malignant) menggunakan pemrosesan PCA dan evaluasi akurasi.