One-Sided Multidimensional Statistical Significance Testing: A New Method of Calculating the Statistical Significance...

Area of Science:

• Statistical analysis
• Spectroscopy
• Nanotechnology

Background:

• Estimating statistical significance between spectra is crucial but existing multivariate tests have limitations.
• Common limitations include one-sided testing, unequal variances, and few repetitions, particularly in magnetic spectroscopy of nanoparticle Brownian motion (MSB).

Purpose of the Study:

• Introduce a novel, powerful, and exact statistical test (T-S test) for comparing spectra.
• Address limitations of existing methods for one-sided testing, unequal variances, and small sample sizes in spectral analysis.

Main Methods:

• The T-S test combines one- or two-sided p-values at each frequency using Stouffer's method.
• Evaluated using simulated and measured magnetic spectroscopy of nanoparticle Brownian motion (MSB) spectra.
• Compared performance against the mean of the spectrum (A-T) for single-sided tests and Hotelling T2 for two-sided tests.

Main Results:

• The T-S test is as powerful as existing methods with uniform variance and outperforms with non-uniform noise power.
• For two-sided tests, the T-S test shows only a slight power decrease with many repetitions but significant outperformance with few repetitions.
• Demonstrated 1 nanogram sensitivity for MSB spectrometer and identified 15 pM concentrations of mouse IL-6, compared to 76 pM by the mean of spectra method.

Conclusions:

• The T-S test provides an exact and powerful method for spectral significance testing, overcoming limitations of current approaches.
• The T-S test significantly enhances sensitivity and detection limits in applications like magnetic spectroscopy of nanoparticle Brownian motion (MSB).
• This new method improves the identification of low concentrations in biological samples.