ErrorByMember: Summarize errors by member
In gerwathome/runOPM: Runs and Examines Reservoir Simulation Models with OPM

Description Usage Arguments Details Value

A member is defined as the data associated with a unique grouping of all cases, well/group names and keywords. Summary statistics are calculated for each member.

1 2	ErrorByMember(long, basedir = "tmp", skip_cases = "HIST", wgnames = NULL, keywords = NULL, errortypes = c("ABS_MEAN_FRAC_ERR"))

`long`	The full path to a .csv, .xls, or .xlsx file containing production history data.
`basedir`	The path to the base directory of a simulation project. The default is a subdirectory of the current directory called "tmp". This is necessary for saving the results.
`skip_cases`	It may be inappropriate to include some cases in the error statistics. The default is to skip all cases with HIST in the name. This parameter may be a character vector and/or a case insensitive regular expression.
`wgnames`	A list with well/group/field names for which errors need to be calculated. Default is all WELL/FIELD/GROUP names in the dataframe.
`keywords`	A list with parameters to be for which errors need to be calculated, e.g. "WOPR". Default is all keywords in the dataframe.
`errortypes`	A list with error types to be calculated. Available error types are c("ABS_MEAN_ERR", "ABS_MEAN_FRAC_ERR", "MAX_ERR", "MAX_FRAC_ERR", "MEAN_ERR", "MEAN_FRAC_ERR", "MIN_ERR", "MIN_FRAC_ERR", "NORM_PROB_ERR", "NORM_PROB_FRAC_ERR", "SLOPE_ERR", "SLOPE_FRAC_ERR") The default is "ABS_MEAN_FRAC_ERR".

The statistics calculated for each error(ERR) and error fraction (FRAC_ERR) are minimum, maximum, mean, mean of absolute values, the approximate probability for the Shapiro-Wilk Normality test, and the slope of the errors with respect to date. The probability associated with the Shapiro-Wilk test is a null hypothesis probability, that is, the smaller it is, the more likely that the tested sample is from a normal distribution. This approach probably isn't as good as looking at the residuals plot, but it may be helpful when trying to examine the errors for a large number of parameters.

The intent of these criteria is to ensure that the quality of the fit to the historical data is comparable to the expected error in a forecast. Unfortunately, trying to optimize multiple criteria is very compute intensive, so this may be a fool's game. Perhaps the best approach will be to pick a single criteria for optimization, and then check the other criteria.

A 'member' is the set of data associated with a particular CASENAME, WGNAME, and KEYWORD. The items within a member each have different dates. The 'member error' is the summary statistic comparing this member to the equivalent data for the base case.

An 'element' is the set of data associated with a particular WGNAME, KEYWORD, and ERRORTYPE. The items within an element each have different casenames. A proxy model created from an element will allow one to estimate the value of an error based on the input values to an experimental design.

Returns a data frame with various summary statistics for each member, and wrties out a csv file in the REPORTS directory.

gerwathome/runOPM documentation built on May 20, 2019, 4:05 p.m.