ELIM/MILE ELIM Ver. 2007.07.18 Procedure for automatic outliers elimination/restoration Description of menu items. 같같같같같같같같같같같같� (H) -- Put on-line help on the screen, which you are reading just now. (A) -- Set acceleration factor ( greater than 0 ). This option allows to speed up the process of elimination/restoration by the expense of some worsening performance: residuals are updated after N operations of elimination/restoration, where N is the acceleration factor. Best choice is N=1, but it may lead to too slow work for session with tens of thousands observations. Increasing acceleration factors decreases computation time approximately by N times, but it may lead not the optimal choice of the outlier/candidate_in_restoration. It is not recommended to use acceleration factor more than 1 unless computational expenses appear really intolerable. (X) -- Set upper limit for formal uncertainty. Observations with formal uncertainty (full sigma: correlator supplied and reweight) exceeding the limit will be marked as outliers in ELIM mode and will not be considered eligible for restoration in MILE mode. If the limit is set to zero then this criteria will not be taken into account. (E) -- EQM speed-up. Toggles flag of whether to save equations of conditions in memory. Saving equations of conditions in memory speeds up computation considerably. However, it may appear that it is not enough operating memory to keep all equations of conditions. Then this flag should be lifted. EQM speed-up flag can be reset only before processing the first observation. (U) -- Set upper threshold (in psec) for detection of outlier. Each observation will be passed through two tests: threshold test and n-sigmas cutoff test. If (unweighted) residual of the observation used in solution exceeds in modulo the absolute value of the specified threshold, this observation is marked as outlier in ELIM mode. If (unweighted) residual of not used in solution but recoverable observation is less in modulo than the absolute value of the specified threshold then this observation is marked as a candidate to restoration in MILE mode. If the value of threshold is set up to zero, then threshold criterion will not be used. The threshold criteria may be negative. The sign of the threshold criteria is used when both the threshold and cutoff criteria are used (see below). (C) -- Set upper level of cutoff limit for detection of outlier using n-sigma criterion. Each observation is passed through two tests: threshold test and n-sigmas cutoff test. If normalized residual of the observation used in solution exceeds in module the specified cutoff limit then this observation is marked as outlier in ELIM mode. If normalized residual of the recoverable but not used in solution observation is less in module than the specified cutoff limit then this observation is marked as a candidate to restoration in MILE mode. If the value of cutoff is set up as zero then n-sigmas cutoff criterion will not be used. If both the threshold and n-sigma criteria are used in ELIM mode and the threshold is positive, then an observation is considered as a candidate for outlier if its residual EITHER greater than the threshold OR its normalized residual greater than the cutoff criteria. The observation with the maximal in modulo normalized residual among those which satisfy either criterion is marked as the best candidate for elimination. If both the threshold and n-sigma criteria are used in ELIM mode and the threshold is negative then an observation is considered as a candidate for outlier if its residual greater than the absolute value of the threshold AND its normalized residual is greater than the cutoff criteria. The observation with the maximal in modulo normalized residual among those which satisfy both criteria is marked as the best candidate for elimination. If both the threshold and n-sigma criteria are used in MILE mode and the threshold is positive, then an observation is considered as a candidate for restoration if its residual is EITHER less than the threshold OR its normalized residual is less than the cutoff criteria. The observation with the minimal in modulo normalized residuals among those which satisfy either criterion is marked as the best candidate for restoration. If both the threshold and n-sigma criteria are used in MILE mode and the threshold is negative, then an observation is considered as a candidate for restoration if its residual is less than the absolute value of the threshold AND its normalized residual is less than the cutoff criteria. The observation with minimal in modulo normalized residuals among those which satisfy both criteria is marked as the best candidate for restoration. (Y) -- Normalization of "normalized postfit residuals" may be done in two modes: 'global' mode when dispersion of postfit residuals used for normalization is calculated for all used observations of the database or in 'baseline' mode when this dispersion is calculated for used observations made at the same baseline as the considered observation. (Q) -- Quality code limit sets the minimal quality code acceptable for the observation considered as candidate in restoration. All observations with quality code less than specified limit are considered as unrecoverable ones and cannot be restored by MILE. Quality code limit may have values in the range [1, 9]. (D) -- Statistics of the solution is not calculated automatically when ELIM is called. Entering D you can initialize ELIM/MILE and update statistics of the solution. (M) -- Try to resolve ambiguity. If YES then MILE tries to resolve ambiguities on the fly. It checks whether the observation is a good candidate for restoration after getting rid from the ambiguities. If the observation with ambiguities is the best candidate for restoration them MILE restores it, changes ambiguities of this observation and possibly changes ambiguities of others suppressed observations of the same scan in order to keep misclosure. MILE will not restore observation if the change of ambiguities with conservation of misclosure would require changes of ambiguities of used observations. (I) -- Ionosphere for ambiguities. Forces MILE to uses effective ionosphere free ambiguities spacing when it tries to resolve ambiguities on the fly. Should be always used when the option (M) is YES except some specific cases. (R) -- Refresh the screen. (-) -- Change singularity check (ref. documentation for SET_SNGCHK). Singularity check action REPARAMETERIZE is recommended to be used in order to avoid abnormal termination caused by normal matrix singularity which may give rise due to elimination too many observations. (') -- Change suppression method. Codes of supported suppression methods will be displayed and used may select appropriate suppression method. Suppression method influence on the status of observations and on possibility to suppress them. (V) -- Verbosity level. Two values are allowed: 0 what means silent mode or 1 -- verbose mode. Some messages will appear at the screen in verbose mode indicating the current work of the program. Only error message will printed out in silent mode. (N) -- Toggle confirmation mode. Confirmation will be inquired before each operation elimination or restoration in confirmation mode. User can reply Y, N, A, S in a response. "Y" means yes, "N" means no, "A" means yes and not to ask further confirmation, "S" means to stop further elimination or restoration and coming to the main menu ELIM. Switching confirmation mode "on" sets automatically verbosity level to 1. (S) -- Save results of the work (to rewrite them from temporary data structures to scratch file), show the listing of parameter estimation by calling CRES and to return to OPTIN. (O) -- Return to OPTIN without saving results. (P) -- Start doing outliers elimination or restoration. (T) -- Toggle mode between ELIM (outliers elimination) and MILE (restoration of suppressed observations). (W) -- Call utility UPWEI (ref. documentation for UPWEI) for updating weights. Additive baseline-dependent (or baseline-independent) corrections to the weights will be found using iterative procedure. After having quadratically added to the a priori weights the ratio of the chi-square to its mathematical expectation become near to unity.