{smcl}
{* 3aug2005/26oct2010}{...}
{hline}
help for {hi:fixsort}
{hline}

{title:Sort variables and align in sorted order, with others fixed in position} 

{p 8 17 2}
{cmd:fixsort}
{it:varlist} 
[{cmd:if} {it:exp}] 
[{cmd:in} {it:range}]
{cmd:,} 
{cmdab:g:enerate(}{it:newvarlist}{cmd:)} 
[ 
{cmdab:miss:ing} 
{cmdab:r:everse} 
]


{title:Description} 

{p 4 4 2} 
{cmd:fixsort} sorts each variable in {it:varlist} and generates new
variables so that all new variables are in sorted order, by default
aligned so that the first value of each is in the first observation, the
second value of each is in the second observation, and so on. Reverse 
sorting (last value first, etc.) is also possible. 

{p 4 4 2}
Any variables not in {it:varlist} retain their current sort order.  
The sort order of the dataset is unaffected. 
Variables may be numeric or string. 


{title:Remarks} 

{p 4 4 2}
{cmd:fixsort} is a rarely needed command. Almost all sorting problems
call for {help sort} or {help gsort}. An example where it may be
appropriate is in sorting independent batches of random numbers held in
separate variables. 

{p 4 4 2}
{cmd:fixsort} is unusual in various ways: 

{p 8 8 2}
Whatever is stored in any observation for new variables has no necessary
relationship to values for other variables in the same observation. 

{p 8 8 2}
By default missing values of any kind (. .a ... .z "") are all ignored.
Any such missing values are represented in new variables by values of .
or "", as appropriate to the variable type, all stored in the later
observations of the dataset. Note that treatment of missing values is
variable by variable, so that the new variables created by {cmd:fixsort}
may readily possess different numbers of non-missing values. See also
the {cmd:missing} option below. 

{p 8 8 2}
{cmd:if} and {cmd:in} act only to select observations and do not affect
the storage of new variables in observations 1 up. 

{p 4 4 2}
By conscious design, there is no {cmd:replace} option to let users overwrite 
existing variables. Thus no misunderstanding of {cmd:fixsort} should mean that 
users lose data. 


{title:Options} 

{p 4 8 2}{cmd:generate()} specifies the names of new variables to hold the 
sorted values of {it:varlist}. {cmd:generate()} is a required option.  

{p 4 8 2}{cmd:missing} specifies that missing values of any kind are to
be included in the sorting. Missing values are treated as is; any other
missing values (corresponding to observations excluded by {cmd:if}
and/or {cmd:in}) are created as "" or . (i.e. system missing). 

{p 4 8 2}{cmd:reverse} specifies sorting in reverse order, i.e. highest 
values first. 
 

{title:Examples} 

{p 4 4 2} 
Suppose we start in each of the following with these data: 

{space 8}{cmd:a   b    c}
{space 8}3   7   13 
{space 8}1   8   12
{space 8}2   9   11 

{p 4 4 2}After {cmd:fixsort a b c, gen(A B C)} we would have 

{space 8}{cmd:a   b    c   A   B    C}
{space 8}3   7   13   1   7   11  
{space 8}1   8   12   2   8   12
{space 8}2   9   11   3   9   13 

{p 4 4 2}After {cmd:fixsort c in 2/3, gen(C23)} we would have 

{space 8}{cmd:a   b    c  C23}
{space 8}3   7   13   11
{space 8}1   8   12   12   
{space 8}2   9   11    . 


{title:Acknowledgments} 

{p 4 4 2}
Philippe Van Kerm provoked a reconsideration and rewriting of this program. 


{title:Author}

{p 4 4 2}Nicholas J. Cox, Durham University, U.K.{break} 
         n.j.cox@durham.ac.uk


{title:Also see}

{p 4 13 2}On-line: help for {help sort}, {help gsort}, {help clsort} (if installed)