Statistical Methods
Check which frequency a sampled sinusoid can imitate
Define the task
Record the original sampling rate and the highest scientifically relevant frequency before reducing a sequence. Merely keeping every second or third observation can make a high-frequency component look like a lower one. Anti-alias filtering is part of a defensible downsampling procedure, and the filter must be chosen before discarded observations are lost.
Carry out the check
For a single real sinusoid sampled at rate fs, frequencies separated by integer multiples of fs can produce equivalent sampled oscillations. Fold a nonnegative input frequency into the interval zero to fs/2 by reducing it modulo fs and taking the smaller of that remainder and fs minus the remainder. This predicts a frequency ambiguity; it does not recover the original signal or specify an adequate filter cutoff and transition band.
Worked hypothetical example
For a hypothetical 70 Hz sinusoid sampled at 100 Hz, the folded frequency is 30 Hz. At 120 Hz sampling it folds to 50 Hz, while sampling above 140 Hz would put 70 Hz below Nyquist. Exactly at Nyquist, phase matters and a sampled sine can disappear, so equality is not a robust design target. The tool checks frequencies only. For an actual downsampling pipeline, document filter design, phase handling, sampling units and the retained frequency band, then validate on synthetic signals.
Method checked 15 November 2026 against SciPy documentation: decimate.