Here’s the upgraded view you asked for — now with that **percent error curve** riding alongside the spectra so you can see the mismatch in both absolute and relative terms.
- **Blue line** → record spectrum values. - **Green line** → target spectrum values. - **Red dashed line with triangles** → percent error at each period, showing whether the record overshoots (+) or undershoots (–) the target.
You’ll notice the red curve spikes to about **–6 %** at 0.40 s (the biggest undershoot) and hovers around ±4 % elsewhere. That’s exactly why the normalized RMSE lands at ~5 % — the deviations are small and fairly balanced.
This kind of overlay is great for quickly spotting *where* the fit is tight and where it drifts, without having to scan numbers in a table. If you’d like, we could also shade the match window (T₁–T₂) in the plot so it’s visually obvious which periods are actually contributing to the score.
"Data speak for themselves" -Reverend Thomas Bayes 1702-1761 P(Ai|E)=(P(E|Ai)P(Ai))/P(E)
