How to deal with missing data

hrkg

Hello,

I need to run a descriptive analysis on a huge number of columns. To do so I made the following script :

dt = Current Data Table();

column_names = dt << Get Column Names(numeric,String);

for(i=3,i<=NItems(column_names),i++,

  colName = column_names;

  cmd = Expr(

  ow = dt << Distribution(

  Continuous Distribution( Column( cName ) ),

  Where( :SEX == "F" )

   ); 

owr = ow<<report;

);

Substitute Into(cmd,Expr(cName),colName);

Eval(cmd);

);

Current Journal() << Save HTML(".....\test.htm");

Now, when I run it I get an error message in log window "One or more columns have nothing but missing values.".

I have some columns that all values or only those for specific "SEX" are missing data. How could I check it in my loop and skip them in analysis ?

Thanks

ms

The stack is just an intermediary step to arrange data in a single column. The error may be that either the label or data columns don't get the names "Position" and "Data" that the summary command assumes here. Take away the "invisible" option and look at the stacked table to see the column names, and change Summary accordingly if something is not named as intended (as noted in the jsl comment above the Stacked Data Column("Data") argument seems to be ignored in JMP 10 if columns are grouped as in your original table, hence the ungrouping command i my example above).

Instead of stack, if that's misbehavin, you can instead use summary and transpose in tandem to get a similar result. Try the example below (works with the posted table "Example") .

dt = Data Table( "Example" );

column_names = dt << Get Column names( numeric );

(dt << Summary(

  Group( :SEX ),

  Mean( Eval( column_names ) ),

  statistics column name format( "column" )

)) << Transpose( Label( :SEX ), Output Table( "Result" ) );