I'm working on ternary KDE plots for my study, and I can't figure out what's going wrong. Hoping someone here can help before I lose another two days (and my sanity) tweaking this! What's Happening? The top panel shows scatter plots (X, Y, and Z) for different parameter combinations (A, B, C, D, and E). These look fine—no issues here. The bottom panel is my attempt at KDE density plots using the same datasets. What I Expected vs. What I Got: X Plot (Leftmost): The dataset is sparse, so I expected the density plot to be very faint or almost nonexistent on a global scale. Y & Z Plots: These have much larger datasets, so I expected clearer and denser KDE distributions, which seem reasonable. However, the density plot for X still appears more structured than I expected. Shouldn't it be almost blank or significantly lower in intensity compared to Y and Z? What I'm Using: Matplotlib + SciPy’s gaussian_kde for density estimation. Barycentric transformation for ternary conversion. I'm still new to KDE and ternary plotting, so any advice would be hugely appreciated before I lose another weekend trying to figure this out! Thanks in advance! For KDE I've been using this script: enter image description here ` I plotted ternary KDE using gaussian_kde from SciPy after transforming my ternary data into Cartesian coordinates. I expected the X plot (leftmost) to have little to no density due to its sparse data, while Y and Z (with more data) should show stronger density gradients. However, the X plot still has a structured KDE pattern instead of being almost blank or very faint. I'm trying to understand if this is due to bandwidth selection, normalization issues, or something else. Continue reading...
