How to change the comparisons of missing data values to behave like prior versions of Statistica?
Article ID: KB0076127
Updated On:
| Products | Versions |
|---|---|
| Spotfire Statistica | 13.5 and higher |
Description
With TIBCO Statistica 13.5 comparisons when missing data (MD) values are involved has changed.
Prior to Statistica version 13.5 if one of the conditions/variables in C1=C2 or C1<>C2 was a NULL/Missing data (MD) then the result would always be NULL/Missing data. While this logic made sense for operations such as >, <,<= or <= , this was not very logical for comparison operators such as = and <>.
Here is a small table that explains the results prior to 13.5 and after 13.5 specific to missing data.
Attached is another summary that specifically addresses how these comparisons have changed while comparing Text Labels to String
Prior to Statistica version 13.5 if one of the conditions/variables in C1=C2 or C1<>C2 was a NULL/Missing data (MD) then the result would always be NULL/Missing data. While this logic made sense for operations such as >, <,<= or <= , this was not very logical for comparison operators such as = and <>.
Here is a small table that explains the results prior to 13.5 and after 13.5 specific to missing data.
| Operation | 13.5 and later | 13.4 and prior |
| If Non-MD=MD | FALSE | NULL/Missing Data |
| If Non-MD<>MD | TRUE | NULL/Missing Data |
| If MD=MD | TRUE | NULL/Missing Data |
| If MD<>MD | FALSE | NULL/Missing Data |
| Operation | 13.5 and later | 13.4 and prior |
|---|---|---|
| If Text(Non-MD)=Label(MD) | FALSE | NULL/Missing Data |
| If Text(Non-MD)<>Label(MD) | TRUE | NULL/Missing Data |
| If Text(MD)=Label(non-MD) | FALSE | FALSE |
| If Text(MD)<>Label(non-MD) | TRUE | TRUE |
| If Text(MD)=Label(MD) | TRUE | NULL/Missing Data |
| If Text(MD)<>Label(MD) | FALSE | NULL/Missing Data |
| If Text(Non-MD) [<] or [<=] Label(MD) | FALSE | NULL/Missing Data |
| If Text(Non-MD) [>] or [>=] Label(MD) | TRUE | NULL/Missing Data |
| If Text(MD) [<] or [<=] Label(non-MD) | TRUE | TRUE |
| If Text(MD) [>] or [>=] Label(non-MD) | FALSE | FALSE |
| If Text(MD) [<] or [>] Label(MD) | FALSE | NULL/Missing Data |
| If Text(MD) [<=] or [>=] Label(MD) | TRUE | NULL/Missing Data |
Please find attached two spreadsheets that illustrate the changes with examples one showing what the comparison operators would return before and one showing the new behavior. Here is a summary:
- = and <> operators will now return true (false) if both operands are MD (instead of returning MD) and will return true (false) if exactly one is MD (as opposed to returning MD).
- <= and >= , in the equality case, they return true for both MD.
- < and >; if both operands are MD, and we say they are equal, then < and > should be false.
- When one value is text, comparison operators will treat MD as empty string and return a valid value
Environment
Windows 8.1, Windows 10, Windows Server 2008 R2, Windows Server 2012 R2, and Windows Server 2016
Resolution
To revert comparisons of Missing Data values to pre-13.5 behavior:
Launch Statistica and navigate to File>>Tools>>Options>>Spreadsheets and enable "Use Legacy MD comparisons":
Close Statistica and re-open for this change to take effect.
To propagate the change to all other users , refer to : https://support.tibco.com/s/article/How-to-propagate-Statistica-options-with-StatOpts-xml-and-DefOpts-xml
Issue/Introduction
This article provides a summary on missing data comparison (MD) and shows the difference between the implementation of missing data comparisons in Statistica 13.5 while compared to the versions prior to 13.5.
Feedback
Yes
No
Powered by
