User manual SAS ANALYTICS ACCELERATOR 1.3 FOR TERADATA

[. . . ] SAS Analytics Accelerator 1. 3 for Teradata Guide ® SAS® Documentation The correct bibliographic citation for this manual is as follows: SAS Institute Inc. SAS® Analytics Accelerator 1. 3 for Teradata: Guide Copyright © 2011, SAS Institute Inc. , Cary, NC, USA All rights reserved. For a hard-copy book: No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, or otherwise, without the prior written permission of the publisher, SAS Institute Inc. For a Web download or e-book: Your use of this publication shall be governed by the terms established by the vendor at the time you acquire this publication. [. . . ] A superscript of 2 (Yes2 or No2 ) indicates an option that works differently for in-database computing compared to out-of-database computing. A superscript of 3 (Yes3 , No3 , N/A3 , or Sometimes3 ) indicates an option that causes the procedure to use SAS/ACCESS for out-of-database computing. A superscript of 4 (Yes4 or No4 ) indicates an option that neither works for in-database computing nor prevents in-database computing, but is unlikely to cause a serious error. A superscript of 5 (Yes5 or No5 ) indicates an option that might cause incorrect answers or Teradata syntax errors. NOTE: As a data set option, DATABASE= is an alias for SCHEMA=, but as LIBNAME options, DATABASE= and SCHEMA= are distinct options. The data set options DATABASE= and SCHEMA= work the same way as the LIBNAME option SCHEMA=. If you specify the same option as both a data set option and as a LIBNAME option, the value of the data set option is used for that data set. Table 1. 5 SAS/ACCESS Data Set Options for Teradata Data Set Option BL_CONTROL= BL_DATAFILE= DATA= N/A N/A OUT= N/A N/A Comments For out-of-database I/O. Chapter 1: SAS Analytics Accelerator 1. 3 for Teradata: Guide Data Set Option BL_LOG= BUFFERS= BULKLOAD= CAST= CAST_OVERHEAD_MAXPERCENT= DATABASE= DATA= N/A N/A N/A N/A N/A No5 OUT= N/A N/A N/A N/A N/A No5 Comments For out-of-database I/O. DATABASE= as a data set option does not work like DATABASE= as a LIBNAME option. For out-of-database I/O. DBCOMMIT= DBCONDITION= DBCREATE_TABLE_OPTS= DBFORCE= N/A No3 N/A N/A N/A No3 Yes1 N/A In-database computation works like DBFORCE=YES, regardless of what value is specified for DBFORCE=. However, DBFORCE= would have an effect on in-database computation only if DBTYPE= specified a Teradata data type that resulted in truncation, whereas specifying DBTYPE= prevents in-database computation. For out-of-database I/O. DBINDEX= DBKEY= N/A N/A N/A N/A SAS/ACCESS Data Set Options for Teradata !33 Data Set Option DBLABEL= DATA= N/A OUT= N/A Comments DBLABEL= applies only when the DATA= data set is a SAS data set with variable labels and the OUT= data set is a DBMS table. In-database computation cannot be performed when the DATA= data set is not a DBMS table. DBMASTER= DBNULL= DBSASLABEL= N/A N/A3 Yes1 N/A No3 N/A DBSASLABEL=NONE is recommended for correct procedure output, but it is normally easier to specify this in the LIBNAME statement. For out-of-database I/O. DBSLICE= DBSLICEPARM= DBTYPE= N/A N/A N/A3 N/A N/A Sometimes3 DBTYPE= prevents in-database computation because some Teradata types such as DATE and TIMESTAMP have not been implemented. For out-of-database I/O. ERRLIMIT= MBUFSIZE= ML_CHECKPOINT= ML_ERROR1= ML_ERROR2= ML_LOG= ML_RESTART= ML_WORK= MULTILOAD= N/A N/A N/A N/A N/A N/A N/A N/A N/A 34 !Chapter 1: SAS Analytics Accelerator 1. 3 for Teradata: Guide Data Set Option MULTISTMT= NULLCHAR= NULLCHARVAL= PRESERVE_COL_NAMES=NO READ_ISOLATION_LEVEL= READ_LOCK_TYPE= READ_MODE_WAIT= SCHEMA= SET= DATA= N/A N/A3 N/A3 No3 No4 No4 No4 Yes1 N/A OUT= N/A No3 No3 No3 No4 No4 No4 Yes1 Yes2 Comments For out-of-database I/O. The default value YES works. The option is ignored. If the DATA= table contains duplicate rows, the procedure attempts to insert duplicate rows into the OUT= table, and the results for in-database and out-of-database processing differ. With in-database processing, duplicate rows are discarded and do not cause errors. With out-of-database processing, duplicate rows cause errors and a ROLLBACK is issued. With MODE=TERADATA, the ROLLBACK produces an empty or incomplete table. The option is ignored. SLEEP= TENACITY= UPDATE_ISOLATION_LEVEL= UPDATE_LOCK_TYPE= UPDATE_MODE_WAIT= N/A N/A No4 No4 No4 N/A N/A No4 No4 No4 Base SAS Data Set Options ! [. . . ] You must be careful when formatting your SQL code for in-database processing. Proc SQL noerrorstop; connect to &dbms. (&connopt. ); execute ( create view tera. tdview_reg as select age, amount, checking, coapp, depends, duration, employed, existcr, 'foreign', history, housing, installp, job, marital, other, property, purpose, resident, savings, telephon, case when good_bad = 'good' then 0 else 1 end as target from tera. dmagecr) by &dbms; execute (commit) by &dbms. Even though the conversion of a binary character dependent vari- able to a numeric variable for use in a linear regression model is not optimal, it is a common operation for using the efficient REG procedure for model selection and exploratory modeling steps. SAS output provides clues about how the factors are used in the model. title3 'Linear Regression on raw data in teradata'; ods select SelectionSummary; proc reg data=tera. tdview_reg; model target = age amount checking coapp depends duration employed existcr history housing installp job marital other property foreign resident savings telephon / selection = stepwise; run; quit; Example ! [. . . ]