Examines and replaces missing values - ------------------------------------
^missing^ varlist ^, an^alysis ^me^thod^(^method-option^)^ [ time-series-option ]
Description - -----------
^missing^ replaces missing values for the variables in ^varlist^
Options - -------
^an^alysis displays a table of association measures. Specifically, Simple and Jaccard coefficients and their significance levels are calculated. High coefficient values correspond to strong relationships between variables.
^me^thod^(^method-option^)^ specifies the method used for replacing missing values in ^varlist^. Available methods are:
^dr^op drops observations for which any variable takes on missing value.
^im^pute makes use of the @impute@ ado file for performing best subset regression. Since regression does not assume causality, each variable is modelled as a combination of the rest.
^in^ter^[^varname^]^ replaces missing values with linear interpolations of the existing values for each group defined by varname. When the missing values are placed at the beginning (end) of the group, the first (last) available value of that group is repeated. Interpolation only makes sense in case of time-series. In order to sort by date the time-series option is required:
^me^an^[^varname^]^ replaces missing values with the mean value for each group defined by varname.
^pr^edict fits (@fit@ command) a linear model to all the variables in varlist a replaces missing values with predicted (@predict@ command) values. Note that this method does not assure that all the missing values are filled in.
Bibliography - ------------
A.K. Jain, R.C. Dubes (1988) "Algorithms for Clustering Data". Prentice Hall.
Examples - --------
. ^missing mpg cx, me(dr)^ . ^missing mpg cx, me(in[maker]) da(year)^ . ^missing mpg cx, me(me[maker])^
Also see - --------
On line help: @fit@, @impute@, @predict@
Author - ------
Jose Maria Sanchez Saez