{smcl}
{* 18-Oct2004}
{hline}
help for {hi:doubletofloat}
{hline}

{title:Change double-type variables to float where possible.}

{p 8 17 2}
{cmd:doubletofloat} [{it:varlist}]

{title:Description}

{p 4 4 2}
This is simply a convenience, issuing {cmd:recast float} {it:varlist2}, where
{it:varlist2} is the set of variables in {it:varlist} that are of type
{cmd:double}.
This has some resemblance to, and is intended to complement {cmd:compress},
which does not include the possibility of changing {cmd:double} to
{cmd:float}.

{p 4 4 2}
{cmd:doubletofloat} would be typically used immediately following
{cmd:compress}, and saves you the trouble of seeking out all the remaining
{cmd:double} variables.  There are some potental complications if you do not
preceed it with {cmd:compress}; see notes under {bf:Remarks}.

{p 4 4 2}
If no {it:varlist} is specified, {cmd:_all} is assumed.

{p 4 4 2}
{cmd:doubletofloat} ignores variables that are not {cmd:double}.


{title:Examples}

{p 4 8 2}{cmd:. doubletofloat}{p_end}

{p 4 8 2}{cmd:. doubletofloat wage income}{p_end}


{title:Remarks}

{p 4 4 2}
This is for those instances when you have some {cmd:double} variables which
could be stored as {cmd:float} without any loss of precision, but do require
{cmd:float} rather than an integer type.
Thus, there are fractional values present, but all can be represented exactly
using {cmd:float}.  Typical values would have fractional parts involving
halves and quarters, etc., such as in 2.5, 2.75, and 10.375.

{p 4 4 2}
It is best to use this immediately following {cmd:compress}.  If you use it
without a preceeding {cmd:compress}, you might recast some variables to {cmd:float}
which would otherwise become {cmd:long}, {cmd:int}, or {cmd:byte}.
If you use it without a preceeding {cmd:compress}, but you do follow it with
{cmd:compress}, then those variables that {cmd:compress} would have recast as
{cmd:int} or {cmd:byte} (but not {cmd:long}) had it been called first,
will still receive that change.

{p 4 4 2}
For a similar related program, see {help floattolong} by the same author.
Both {cmd:doubletofloat} and {cmd:floattolong} are intended as companions to
{cmd:compress}, and would best be used as follows:

{p 4 8 2}{cmd:. compress}{p_end}
{p 4 8 2}{cmd:. floattolong}{p_end}
{p 4 8 2}{cmd:. doubletofloat}{p_end}

{p 4 4 2}
Note that in this sequence, the relative placement of {cmd:floattolong} is
not critical, as long as {cmd:compress} comes before {cmd:doubletofloat}.
(The sequence {cmd:doubletofloat}, {cmd:compress}, {cmd:floattolong} should
also work properly, but it is easiest to remember to place {cmd:compress}
first.)

{p 4 4 2}
In general data preparation, {cmd:doubletofloat} would usually be called
without a {it:varlist}; there is little point in calling it with a specific
set of variables, since you could just as well type {cmd:recast float} 
with the same {it:varlist}.  The exception is where you use it within a
program that generates a variable that is initially {cmd:double}, and you
want to clean up that one variable without affecting others.


{title:Author}

{p 4 4 2}
David Kantor, Institute for Policy Studies, Johns Hopkins University.
Email {browse "mailto:dkantor@jhu.edu":dkantor@jhu.edu} if you observe any
problems.

{title:Also see}
{help compress}, {help recast}, {help floattolong}, {help datatypes}