我有几个问题请教，原文如下：    FFT analysis of the (gene) expression profiles.     Fourier analysis was performed on each profile in the quality-controlled set (5,081 oligonucleotides). Profiles were smoothed with missing values imputed using a locally weighted regression algorithm with local weighting restricted to 12% using R. Fourier analysis was performed on each profile using the fft() function of R, padded with zeros to 64 measurements. The power spectrum was calculated using the spectrum() function of R. The power at each frequency (Power()), the total power (Ptot), and the frequency of maximum power (Fmax) were determined. The periodicity score was defined as Power[(Fmax-1) + (Fmax) + (Fmax+1)]/Ptot. The most frequent value of Fmax across all profiles was deemed the major frequency (m) and used in determining phase information. The phase of each profile was calculated as atan2[-(I (m)],R (m), where atan2 is R’s arctangent function and I and R are the imaginary and real parts of the FFT. Profiles were then ordered in increasing phase from -p to p. The loess smooth profiles were drawn through the raw expression data using the loess() function found in the modern regression library of R (version 1.5.1). The default parameters were used, with the exception that local weighting was reduced to 30%. For the averaged profiles of the functional groups, the loess smooth profiles were calculated for each expression profile individually and subsequently averaged to create the representative profile. These same methods were applied to both the randomized set and the yeast cell cycle dataset.    因为我又相似的数据要处理，但是我在重复该方法时有些地方不明白，在给作者写信未果的情况下，到论坛来求教。    1. Each profile(基因表达谱) was smoothed with missing values imputed using a local weighted regression algorithm...这里不知道怎么做了，基因表达谱是先定义成时间序列的，中间有时间点没有数据，我不知道local weighted regression和restricted by 12% 怎么调用。    2. FFT()中 padded with zero to 64 measurement 怎么做到？    3. periodic score 怎么做出来的？    4. ordered from phase -p to p又是如何实现的？感谢指教和任何有益的建议，盼复。