Inefficiency of SAS Programming

If anyone wants to see a prime example of how inefficient it is to program
in SAS, take a look at the SAS programs provided by the US Agency for
Healthcare Research and Quality for risk adjusting and reporting for
hospital outcomes at http://www.qualityindicators.ahrq.gov/software.htm .
?The PSSASP3.SAS program is a prime example. ?Look at how you do a vector
product in the SAS macro language to evaluate predictions from a logistic
regression model. ?I estimate that using R would easily cut the programming
time of this set of programs by a factor of 4.

Frank
--
Frank E Harrell Jr ? Professor and Chair ? ? ? ? ? School of Medicine
? ? ? ? ? ? ? ? ? ? Department of Biostatistics ? Vanderbilt University

If anyone wants to see a prime example of how inefficient it is to program
in SAS, take a look at the SAS programs provided by the US Agency for
Healthcare Research and Quality for risk adjusting and reporting for
hospital outcomes at http://www.qualityindicators.ahrq.gov/software.htm .
?The PSSASP3.SAS program is a prime example. ?Look at how you do a vector
product in the SAS macro language to evaluate predictions from a logistic
regression model. ?I estimate that using R would easily cut the programming
time of this set of programs by a factor of 4.

2009/2/26 Frank E Harrell Jr <f.harrell at vanderbilt.edu>:

If anyone wants to see a prime example of how inefficient it is to program
in SAS, take a look at the SAS programs provided by the US Agency for
Healthcare Research and Quality for risk adjusting and reporting for
hospital outcomes at http://www.qualityindicators.ahrq.gov/software.htm .
 The PSSASP3.SAS program is a prime example.  Look at how you do a vector
product in the SAS macro language to evaluate predictions from a logistic
regression model.  I estimate that using R would easily cut the programming
time of this set of programs by a factor of 4.

 Plenty of examples ripe for sending to www.thedailywtf.com there. Like this:

    IF &N. =  1 THEN SUB_N = 1;
    IF &N. =  3 THEN SUB_N = 2;
    IF &N. =  4 THEN SUB_N = 3;
    IF &N. =  6 THEN SUB_N = 4;
    IF &N. =  7 THEN SUB_N = 5;
    IF &N. =  8 THEN SUB_N = 6;
    IF &N. =  9 THEN SUB_N = 7;
    IF &N. = 10 THEN SUB_N = 8;
    IF &N. = 11 THEN SUB_N = 9;
    IF &N. = 12 THEN SUB_N = 10;
    IF &N. = 13 THEN SUB_N = 11;
    IF &N. = 14 THEN SUB_N = 12;
    IF &N. = 15 THEN SUB_N = 13;
    IF &N. = 17 THEN SUB_N = 14;
    IF &N. = 18 THEN SUB_N = 15;
    IF &N. = 19 THEN SUB_N = 16;

Of course it's possible to write code like that in any language, it
just looks worse when it's in ALL CAPS and written in a style that
looks like the 1980s and onward never happened. The question is
whether it's possible to write this better in SAS. Most of us on this
list could write it in R in a better way.

2009/2/26 Frank E Harrell Jr <f.harrell at vanderbilt.edu>:

If anyone wants to see a prime example of how inefficient it is to program
in SAS, take a look at the SAS programs provided by the US Agency for
Healthcare Research and Quality for risk adjusting and reporting for
hospital outcomes at http://www.qualityindicators.ahrq.gov/software.htm .
?The PSSASP3.SAS program is a prime example. ?Look at how you do a vector
product in the SAS macro language to evaluate predictions from a logistic
regression model. ?I estimate that using R would easily cut the programming
time of this set of programs by a factor of 4.

 Plenty of examples ripe for sending to www.thedailywtf.com there. Like this:

    IF &N. =  1 THEN SUB_N = 1;
    IF &N. =  3 THEN SUB_N = 2;
    IF &N. =  4 THEN SUB_N = 3;
    IF &N. =  6 THEN SUB_N = 4;
    IF &N. =  7 THEN SUB_N = 5;
    IF &N. =  8 THEN SUB_N = 6;
    IF &N. =  9 THEN SUB_N = 7;
    IF &N. = 10 THEN SUB_N = 8;
    IF &N. = 11 THEN SUB_N = 9;
    IF &N. = 12 THEN SUB_N = 10;
    IF &N. = 13 THEN SUB_N = 11;
    IF &N. = 14 THEN SUB_N = 12;
    IF &N. = 15 THEN SUB_N = 13;
    IF &N. = 17 THEN SUB_N = 14;
    IF &N. = 18 THEN SUB_N = 15;
    IF &N. = 19 THEN SUB_N = 16;

Of course it's possible to write code like that in any language, it
just looks worse when it's in ALL CAPS and written in a style that
looks like the 1980s and onward never happened. The question is
whether it's possible to write this better in SAS. Most of us on this
list could write it in R in a better way.

 Barry

If anyone wants to see a prime example of how inefficient it is to program
in SAS, take a look at the SAS programs provided by the US Agency for
Healthcare Research and Quality for risk adjusting and reporting for
hospital outcomes at http://www.qualityindicators.ahrq.gov/software.htm .
?The PSSASP3.SAS program is a prime example. ?Look at how you do a vector
product in the SAS macro language to evaluate predictions from a logistic
regression model. ?I estimate that using R would easily cut the programming
time of this set of programs by a factor of 4.

Frank
--
Frank E Harrell Jr ? Professor and Chair ? ? ? ? ? School of Medicine
? ? ? ? ? ? ? ? ? ? Department of Biostatistics ? Vanderbilt University

Presumably, something like

? ? IF &N. = ?1 THEN SUB_N = 1;
? ? ELSE IF &N. < 5 THEN SUB_N = &N.-1;
? ? ELSE IF &N. < 16 THEN SUB_N = &N.-2;
? ? ELSE SUB_N = &N.-3;

would work, provided that 2, 5, 16 are impossible values. Problem is that it
actually makes the code harder to grasp, so experienced SAS programmers go
for the dumb but readable code like the above.

In R, the cleanest I can think of is

subn <- match(n, setdiff(1:19, c(2,5,16)))

or maybe just

subn <- match(n, c(1, 3:4, 6:15, 17:19))

although

subn <- factor(n, levels = c(1, 3:4, 6:15, 17:19))

might be what is really wanted

Thanks for pointing me to the SAS code, Dr Harrell
After reading codes, I have to say that the inefficiency is not
related to SAS language itself but the SAS programmer. An experienced
SAS programmer won't use much of hard-coding, very adhoc and difficult
to maintain.
I agree with you that in the SAS code, it is a little too much to
evaluate predictions. such complex data step actually can be replaced
by simpler iml code.

On Thu, Feb 26, 2009 at 5:57 PM, Frank E Harrell Jr
<f.harrell at vanderbilt.edu> wrote:

If anyone wants to see a prime example of how inefficient it is to program
in SAS, take a look at the SAS programs provided by the US Agency for
Healthcare Research and Quality for risk adjusting and reporting for
hospital outcomes at http://www.qualityindicators.ahrq.gov/software.htm .
 The PSSASP3.SAS program is a prime example.  Look at how you do a vector
product in the SAS macro language to evaluate predictions from a logistic
regression model.  I estimate that using R would easily cut the programming
time of this set of programs by a factor of 4.

Frank
--
Frank E Harrell Jr   Professor and Chair           School of Medicine
                    Department of Biostatistics   Vanderbilt University

Sometimes for the sake of simplicity, SAS coding is created like that. 
One can use the concatenate function and drag and drop in an simple 
excel sheet for creating elaborate SAS code like the one mentioned and 
without any time at all.

There are multiple ways to do this in SAS , much better and similarly in 
R....

There are many areas that SAS programmers would find R a bit not so 
useful ---example

the equivalence of proc logistic for creating a logistic model.

On Fri, Feb 27, 2009 at 10:21 AM, Wensui Liu <liuwensui at gmail.com 
<mailto:liuwensui at gmail.com>> wrote:

    Thanks for pointing me to the SAS code, Dr Harrell
    After reading codes, I have to say that the inefficiency is not
    related to SAS language itself but the SAS programmer. An experienced
    SAS programmer won't use much of hard-coding, very adhoc and difficult
    to maintain.
    I agree with you that in the SAS code, it is a little too much to
    evaluate predictions. such complex data step actually can be replaced
    by simpler iml code.

    On Thu, Feb 26, 2009 at 5:57 PM, Frank E Harrell Jr
    <f.harrell at vanderbilt.edu <mailto:f.harrell at vanderbilt.edu>> wrote:

     > If anyone wants to see a prime example of how inefficient it is

    to program

     > in SAS, take a look at the SAS programs provided by the US Agency for
     > Healthcare Research and Quality for risk adjusting and reporting for
     > hospital outcomes at

    http://www.qualityindicators.ahrq.gov/software.htm .

     >  The PSSASP3.SAS program is a prime example.  Look at how you do

    a vector

     > product in the SAS macro language to evaluate predictions from a

    logistic

     > regression model.  I estimate that using R would easily cut the

    programming

     > time of this set of programs by a factor of 4.
     >
     > Frank
     > --
     > Frank E Harrell Jr   Professor and Chair           School of Medicine
     >                     Department of Biostatistics   Vanderbilt

    University

     >
     > ______________________________________________
     > R-help at r-project.org <mailto:R-help at r-project.org> mailing list
     > https://stat.ethz.ch/mailman/listinfo/r-help
     > PLEASE do read the posting guide

Frank,

I can't see the code you mention - Web marshall at work - but I don't think
you should be too quick to run down SAS - it's a powerful and flexible
language but unfortunately very expensive.

Your example mentions doing a vector product in the macro language - this
only suggest to me that those people writing the code need a crash course
in SAS/IML (the matrix language). SAS is designed to work on records and so
is inapproprorriate for matrices - macros are only an efficient code
copying device. Doing matrix computations in this way is pretty mad and the
code would be impossible never mind the memory problems.
SAS recognise that but a lot of SAS users remain familiar with IML.

In IML by contrast there are inner, cross and outer products and a raft of
other useful methods for matrix work that R users would be familiar with.
OLS for example is one line:

b = solve(X`X, X`y) ;
rss = sqrt(ssq(y - Xb)) ;

And to give you a flavour of IML's capabilities I implemented a SAS version
of the MARS program in it about 6 or 7 years ago.
BTW SPSS also has a matrix language.

Gerard

                                                                           
             Frank E Harrell                                               
             Jr                                                            
             <f.harrell at vander                                          To 
             bilt.edu>                 R list <r-help at stat.math.ethz.ch>   
             Sent by:                                                   cc 
             r-help-bounces at r-                                             
             project.org                                           Subject 
                                       [R] Inefficiency of SAS Programming 
                                                                           
             26/02/2009 22:57                                              
                                                                           
                                                                           
                                                                           
                                                                           




If anyone wants to see a prime example of how inefficient it is to
program in SAS, take a look at the SAS programs provided by the US
Agency for Healthcare Research and Quality for risk adjusting and
reporting for hospital outcomes at
http://www.qualityindicators.ahrq.gov/software.htm .  The PSSASP3.SAS
program is a prime example.  Look at how you do a vector product in the
SAS macro language to evaluate predictions from a logistic regression
model.  I estimate that using R would easily cut the programming time of
this set of programs by a factor of 4.

Frank
--
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University

I would like to know if we can create a package in which r functions are 
renamed closer to sas language.doing so will help people familiar to SAS 
to straight away take to R for their work,thus decreasing the threshold 
for acceptance - and then get into deeper understanding later.

since it is a package it would be optional only for people wanting to 
try out R from SAS.. Do we have such a package right now..it basically 
masks R functions to the equivalent function in another language just 
for user ease /beginners

for example

creating function for means 

 procmeans<-function(x,y)
+ {
summary (
subset(x,select=c(x,y))
+
)

creating function for importing csv

procimport <-function(x,y)
+ {
read.csv(
textConnection(x),row.names=y,na.strings="  "
+
)


creating function fo describing data

procunivariate<-function(x)
+ {
summary(x)
+
)

regards,

ajay

www.decisionstats.com <http://www.decisionstats.com>

On Fri, Feb 27, 2009 at 4:27 AM, Frank E Harrell Jr 
<f.harrell at vanderbilt.edu <mailto:f.harrell at vanderbilt.edu>> wrote:

    If anyone wants to see a prime example of how inefficient it is to
    program in SAS, take a look at the SAS programs provided by the US
    Agency for Healthcare Research and Quality for risk adjusting and
    reporting for hospital outcomes at
    http://www.qualityindicators.ahrq.gov/software.htm .  The
    PSSASP3.SAS program is a prime example.  Look at how you do a vector
    product in the SAS macro language to evaluate predictions from a
    logistic regression model.  I estimate that using R would easily cut
    the programming time of this set of programs by a factor of 4.

    Frank
    -- 
    Frank E Harrell Jr   Professor and Chair           School of Medicine
                        Department of Biostatistics   Vanderbilt University

Frank,

I can't see the code you mention - Web marshall at work - but I don't

you should be too quick to run down SAS - it's a powerful and flexible
language but unfortunately very expensive.

Your example mentions doing a vector product in the macro language - this
only suggest to me that those people writing the code need a crash course
in SAS/IML (the matrix language). SAS is designed to work on records and

is inapproprorriate for matrices - macros are only an efficient code
copying device. Doing matrix computations in this way is pretty mad and

code would be impossible never mind the memory problems.
SAS recognise that but a lot of SAS users remain familiar with IML.

In IML by contrast there are inner, cross and outer products and a raft

other useful methods for matrix work that R users would be familiar with.
OLS for example is one line:

b = solve(X`X, X`y) ;
rss = sqrt(ssq(y - Xb)) ;

And to give you a flavour of IML's capabilities I implemented a SAS

of the MARS program in it about 6 or 7 years ago.
BTW SPSS also has a matrix language.

Gerard

             Frank E Harrell

             Jr

             <f.harrell at vander

             bilt.edu>                 R list <r-help at stat.math.ethz.ch>

             Sent by:

             r-help-bounces at r-

             project.org

                                       [R] Inefficiency of SAS

             26/02/2009 22:57

If anyone wants to see a prime example of how inefficient it is to
program in SAS, take a look at the SAS programs provided by the US
Agency for Healthcare Research and Quality for risk adjusting and
reporting for hospital outcomes at
http://www.qualityindicators.ahrq.gov/software.htm .  The PSSASP3.SAS
program is a prime example.  Look at how you do a vector product in the
SAS macro language to evaluate predictions from a logistic regression
model.  I estimate that using R would easily cut the programming time of
this set of programs by a factor of 4.

Frank
--
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University

The information transmitted is intended only for the person or entity to

Should you consider that the material contained in this message is

Is le haghaidh an duine n? an eintitis ar a bhfuil s? d?rithe, agus le

M?s rud ? go measann t? gur ?bhar col?il at? san ?bhar at? sa

Ajay ohri wrote:

I would like to know if we can create a package in which r functions
are renamed closer to sas language.doing so will help people familiar
to SAS to straight away take to R for their work,thus decreasing the
threshold for acceptance - and then get into deeper understanding later.

since it is a package it would be optional only for people wanting to
try out R from SAS.. Do we have such a package right now..it basically
masks R functions to the equivalent function in another language just
for user ease /beginners

for example

creating function for means
 procmeans<-function(x,y)
+ {
summary (
subset(x,select=c(x,y))
+
)

creating function for importing csv

procimport <-function(x,y)
+ {
read.csv(
textConnection(x),row.names=y,na.strings="  "
+
)


creating function fo describing data

procunivariate<-function(x)
+ {
summary(x)
+
)

regards,

ajay

Ajay,

This will generate major confusion among users of all types and be hard
to maintain.  A better approach is to get Bob Muenchen's excellent book
and keep it nearby.

Frank

Wensui Liu wrote:

Thanks for pointing me to the SAS code, Dr Harrell
After reading codes, I have to say that the inefficiency is not
related to SAS language itself but the SAS programmer. An experienced
SAS programmer won't use much of hard-coding, very adhoc and difficult
to maintain.
I agree with you that in the SAS code, it is a little too much to
evaluate predictions. such complex data step actually can be replaced
by simpler iml code.

Agreed that the SAS code could have been much better. ?I programmed in SAS
for 23 years and would have done it much differently. ?But you will find
that the most elegant SAS program re-write will still be a far cry from the
elegance of R.

Frank

On Thu, Feb 26, 2009 at 5:57 PM, Frank E Harrell Jr
<f.harrell at vanderbilt.edu> wrote:

If anyone wants to see a prime example of how inefficient it is to
program
in SAS, take a look at the SAS programs provided by the US Agency for
Healthcare Research and Quality for risk adjusting and reporting for
hospital outcomes at http://www.qualityindicators.ahrq.gov/software.htm .
?The PSSASP3.SAS program is a prime example. ?Look at how you do a vector
product in the SAS macro language to evaluate predictions from a logistic
regression model. ?I estimate that using R would easily cut the
programming
time of this set of programs by a factor of 4.

Frank
--
Frank E Harrell Jr ? Professor and Chair ? ? ? ? ? School of Medicine
? ? ? ? ? ? ? ? ? ?Department of Biostatistics ? Vanderbilt University

--
Frank E Harrell Jr ? Professor and Chair ? ? ? ? ? School of Medicine
? ? ? ? ? ? ? ? ? ? Department of Biostatistics ? Vanderbilt University

Immersion therapy can be done at a later stage after the 
newly baptized R  corporate user is happy with the fact that he can do 
most of his legacy code in R easily now .

 I have treading water in the immersion for over a year now.


 Most SAS consultants and corporate users are eager to try out R ..but 
they are scared of immersion especially in these cut back times  ...so 
this could be a middle step...let me go ahead and create the wrapper SAS 
package as a middle ware between r and sas ..

and we will let the invisible hands of  free market decide :))

regards,

ajay

www.decisionstats.com <http://www.decisionstats.com>

I am not a Marxist. 
Karl Marx <http://www.brainyquote.com/quotes/quotes/k/karlmarx131048.html> 

On Fri, Feb 27, 2009 at 8:01 PM, Marc Schwartz 
<marc_schwartz at comcast.net <mailto:marc_schwartz at comcast.net>> wrote:

    on 02/27/2009 07:57 AM Frank E Harrell Jr wrote:

     > Ajay ohri wrote:

     >>
     >> I would like to know if we can create a package in which r functions
     >> are renamed closer to sas language.doing so will help people

    familiar

     >> to SAS to straight away take to R for their work,thus decreasing the
     >> threshold for acceptance - and then get into deeper

    understanding later.

     >>
     >> since it is a package it would be optional only for people

    wanting to

     >> try out R from SAS.. Do we have such a package right now..it

    basically

     >> masks R functions to the equivalent function in another language

    just

     >> for user ease /beginners
     >>
     >> for example
     >>
     >> creating function for means
     >>  procmeans<-function(x,y)
     >> + {
     >> summary (
     >> subset(x,select=c(x,y))
     >> +
     >> )
     >>
     >> creating function for importing csv
     >>
     >> procimport <-function(x,y)
     >> + {
     >> read.csv(
     >> textConnection(x),row.names=y,na.strings="  "
     >> +
     >> )
     >>
     >>
     >> creating function fo describing data
     >>
     >> procunivariate<-function(x)
     >> + {
     >> summary(x)
     >> +
     >> )
     >>
     >> regards,
     >>
     >> ajay

     >
     > Ajay,
     >
     > This will generate major confusion among users of all types and

    be hard

     > to maintain.  A better approach is to get Bob Muenchen's

    excellent book

     > and keep it nearby.
     >
     > Frank

    I whole heartedly agree with Frank here. It may be one thing to have a
    "translation" process in place based upon some form of logical mapping
    between the two languages (as Bob's book provides). But is another thing
    entirely to actually start writing functions that provide wrappers
    modeled on SAS based PROCs.

    If you do this, then you only serve to obfuscate the fundamental
    philosophical and functional differences between the two languages and
    doom a new useR to missing all of R's benefits. They will continue to
    try to figure out how to use R based upon their "SAS intuition" rather
    than developing a new set of coding and even statistical paradigms.

    Having been through the SAS to S/R transition myself, having used SAS
    for much of the 90's and now having used R for over 7 years, I can speak
    from personal experience and state that the only way to achieve the
    requisite proficiency with R is immersion therapy.

    Regards,

    Marc Schwartz

Ajay ohri wrote:

Sometimes for the sake of simplicity, SAS coding is created like that. One
can use the concatenate function and drag and drop in an simple excel sheet
for creating elaborate SAS code like the one mentioned and without any time
at all.

A system that requires Excel for its success is not a complete system.