Data transformations

[MDbiologist]

Employer has gone over/pushing SAS EG 4.1.

All of my experience is with V8 and 9 and code writing.

I'm trying to run a few transformations on a set of data (arcsin, log, log(x+1) etc) but how in the heck do I do it. I don't really see an appropriate place to enter in custom code to transform my variables from the original data set or to tell a statistical test to use transformed variables. Message was edited by: MDbiologist

[deleted_user]

You do it as you have done in the past. Once you have run the nodes to retrieve your analytical data, add a code node to the project.

Then write your transformation code in the code node.

[deleted_user]

1) Every user of EG has to bring a dataset into a project to use it.

2) Run a query against this dataset, usually to select out the specific data you wan to process, and create some "Computed Columns" that do these simple transforms.

One of my computed columns is a parsing of a text field, taking "Thu Feb 14 16:45:13 2008" and producting a usable datetime field = datetimestamp. Along the way I also create datestamp and timestamp.

One of my computed columns recalculates the transaction response time because the source data limited itself to milliseconds and I need to resolve microseconds. So I have computed column called "response_time", even though the source dataset has a column called "response_time" and the output dataset contains the calculated one becuase I put the calculated one in the "Select Data" list and not the originals source "response_time".

Now, when I "Graph" a "Line Plot" of the resulting dataset (from the "Filter and Query ..." ) I put datetimestamp in the horizontal, and response_time in the vertical thingy in the Line Plot wizard/dialog.

Similarly, I needed to do so some correlative analysis, so I ran a "Filter and Query ..." and created square roots, squares and logarithms of the analysis variables and then ran the "Analyze : Multivariate : Correlation" wizard/dialog and had it run the correlations between all the computed columns and the base data and also print the scatter plots. I was then able to inspect the results so that I could choose when variables to use in the regression and ETS time series forecasting analyses.

All compliments of "Computed Columns" and max, min, mean, count and sum summaries from within the Filter and Query ... wizard/dialog box.

[Doc_Duke]

Matt,

With your experience in SAS coding, you might want to get a couple of Books-by-Users. "SAS for Dummies" is a quick read and shows how the pieces of SAS are wrapped up in EGuide. "SAS Programming for Enterprise Guide Users" shows how to incorporate the coding tools you know into the EGuide environment.

EGuide does it's basic data transformations using PROC SQL. If you've never used SQL (and many scientist users haven't), it's worth learning.

Doc