Sam Winstanley’s Blog

About a month ago, SPSS released Dimensions 5.5 and it took me a while to get around to installing it. Usually the .5 releases that have come in Dimensions have not had too many new features but have rolled up patches to the .0 releases. However in the case of Desktop Reporter there are some actual new features that have come through that are worth talking about.

I’ve spent the past week or so testing all of the different forget data tools and solutions with this new version which hasn’t turned up any difficult compatibility issues.

The user interface, is the same great interface we’ve all come to know and love, there’s really no significant changes since version 5 (which itself didn’t change very much since 4.5).

The new features fall into 3 main areas:

• Auto categorization of text, numeric and date variables
• A new netting feature.
• Significantly enhanced data management and data exporting features.
• One last, really important thing at the end.

Auto Categorization

Reporter and its predecessor mrTables have always supported some level of analysis on Text, Date and Numeric variables, for instance you could always create a report of Mean of Age, Max of Age etc. and you have always had the ability to create derived variables with custom elements in them. However its a lot of work to do this…

If you open the SPSS sample database called “Household” there is a variable called Vehicle.Color which is a text variable with a number of distict values as follows:

Up until now, you have had a number of approaches available to deal with these variables:

1. Write a new categorical version of the variable with all of the right categories in it then write some script (either DMOM or SQL) to populate that variable with the values.
2. Use one of the scripts from the DDL to attempt to auto-generate an axis expression for the variable that can be added to the variable.
3. Manually set up User Defined Categories for the variable in reporter.
4. Have somebody manually code the variable.

The first 2 options really only exist if you have a DP process in place for doing this kind of work. Additionally options 1 and 3 both create quite a bit of potential maintenance if you have variables which can hold a lot of values.

If you only use Reporter, then your only option was option 3 and that involved a lot of repetative work manually creating each category.

The new way is very much smarter.. you can simply right-click on a variable and select the new Categorize option:

Which opens up a new screen called “Simple Categorization”:

By default, this screen creates a .Codes variable for the main variable (this is also known as a Helper field), this is the place that special responses such as DK, NA, REF get stored when they exist on a Text/Numeric/Date variable.

The Preview screen is very useful here and shows you what you are going to get before actually generating the new variable:

After going through this process what you end up with looks like this:

And this new variable can be used for reporting in the way you would expect. Pretty good eh!

but what’s it actually done
Some of you might be wondering what’s happened under the covers here. Well to find this out we’ll export the Metadata and have a look at the new variable.

Looking into this exported metadata the Color variable now looks as follows:
color “Color”
text
helperfields (
“Codes” “Codes”
[
IsUserDefined = true
]
categorical [0..1]
{
Black “Black”,
Blue “Blue”,
Green “Green”,
Pink “Pink”,
Purple “Purple”,
Red “Red”,
Red#20and#20blue “Red and blue”,
Red#20and#20yellow “Red and yellow”,
Silver “Silver”,
Too#20rusty#20to#20tell “Too rusty to tell”,
White “White”,
Yellow “Yellow”,
“other” “Other”
} expression(”Categorize(color,”"color.Codes”")”, NoDeriveElements) usagetype(”HelperField”)
helperfields (
“other” “Other”
[
IsUserDefined = true
]
text;
);
);

The sharp eyed will see… a new Function in the DataModel, this is not some kind of Desktop Reporter faking there’s new functionality that you could get at in script here too. The Categorize function itself does not seem to be documented, but there is a new funtionality in the Tables Object Model which you will find in the version 5.5 DDL (http://www.spss.com/ddl) under CreateCategorizedVariable. Generally all the function library functions do get documented so its probably only a matter of time… or perhaps I just failed to find the documentation in the DDL.

Overall, this is a great new feature and there’s a lot of times that this will save time, I see it being particularly useful for working with variables that are coming in from Sample that are often text based.

A new netting feature.

There are plenty of times when you would like to create a Nett of some categories but you have some items which can fall into multiple Netts in the variable. For instance in a list of brands you might want to create a Nett called “Market Leaders” and another Nett called “International Brands” and some categories would fall into both netts.. The new feature is called “Nett and Keep” which SPSS calls “Net and Keep” and it creates a Nett of some elements but leaves the elements in place much like the old “Combine and Keep” feature.

This is useful, its not a show stopping feature.

Data Exporting updates..

Release by release, Desktop Reporter seems to get more and more viable as a DP tool rather than just a reporting tool, possibly much to the horror of DP people. All of this is via the feature “Export Data” which in allows a reporter user to create a custom data extraction which has the following features:
- Filters cases according to desktop reporter features.
- Allows a subset of variables to be extracted.
- Includes newly created derived variables and in the resulting data and any axis expressions.

It’s not completely comprehensive but its very usable and with this release of reporter there’s the following enhancements:

• Export levels data to SPSS files (to be honest, I had problems getting this to work as I would expect with the household data, I only got one output file).
  
• Export Factors - Up until now, when you exported data from Desktop Reporter, the values of categorical variables always were the real, datamodel values. If you are exporting to a native dimensions format its fine, but if you export to CSV its a real pain.. Lets say you have a scale question like
  Extremely Happy(5), Quite Happy(4), Neither (3), Quite Unhappy(2), Extremely Unhappy (1). In CSV you might very well end up with a variable that has values like 324, 325,326,327,328 and these values were unusable without metadata. With this new feature the Factor values (or scores) get exported and this is much more usable.
• New Data Format - DZF - Compressed Dimensions File. Many of you have probably noticed that DDF files are quite big on disk, but they compress very small. In my test a small export was 160kb uncompressed and 2kb zipped. This is not a real new format, its just compressing an existing one, in fact if you access a DZF file it becomes a DDF.. its useful for emailing data though.

One last thing

When you save an MDD file with version 5.5 of Dimensions you cannot access it using version 5.0 of Dimensions. I am not aware of any fundamental compatibilty issues but its a fairly significant irritation if you are preparing data for people and you don’t know which version of Reporter they use. Opening 5.5 data with 4.5 is reported to be okay, it might be a 5.0 glitch and now I’ve upgraded its not that easy for me to see if there is a hack to get around it….you have been warned.