The frame-normalized principal component eigenvalues (left column) and eigenimages (right column) compared to the raw light curve and point-spread function (top row) for visit 1 when not trimming any data (left panel) and when trimming the first 500 integrations (right panel). The components found after trimming the first 500 integrations are used in our light-curve fitting. We include 40 binned data points to aid in visualization of the eigenvalue time series.