NAME: REMOVE_OUTLIERS PURPOSE: Routine eliminates outliers. Quadratic trend is determined in a hollow vicinity of each point. The data value is compared with the trend value. If the deviation is statistically improbable, the value is repaired. There are 6 options for repair to be set in the subroutine remove_outliers_repair.pro. Routine gives the summary of its work: how many of the total number of numeric values were repaired, and the number of failure cases (when it was impossible to establish a trend). CATEGORY: Data Processing CALLING SEQUENCE: remove_outliers, epoch, valuesin, d, tmax, nmax INPUTS: EPOCH: time array for the data values. Any time units may be used, just do it consistently. Double 1D array. VALUESIN: 1D array of values to filter; its numerical values are replaced by filtered data at the end. D: half-size of the hollow vicinity of the point where trend is established (integer) TMAX: maximal time interval covered by the hollow vicinity (double) NMAX: maximal deviation from the trend deemed to be probable (in units of standard deviation). Integer. KEYWORDS: None PARAMETERS: Repair option set in subroutine remove_outliers_repair.pro. OUTPUTS: VALUESIN: Array of filtered values (numerical values of input are replaced). The code may produce "division by zero" warnings originated in the svdfit routine. They should be ignored. DEPENDENCIES: remove_outliers_repair.pro MODIFICATION HISTORY: Written by: Vladimir Kondratovich 2007/12/28.