Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Automating back-transforming variables

From   Eduardo Nunez <>
Subject   st: Automating back-transforming variables
Date   Fri, 20 Aug 2010 16:20:53 -0400

Dear Statalist,

I have a dataset  that needs to be imputed (I used ice) and has many
continuous variables. Because none of them achieved the normality
assumption, I decided to log-transform them before imputation. Once
the dataset was imputed, I need those variables back-transformed to
the scale they were initially.
This is my code:
// looking for the most appropriate transformation

. qladder ca125
. ladder ca125
. lnskew0 lnca125= ca125

. qladder bnp
. ladder bnp
. lnskew0 lnbnp =bnp

Imputation with ice [include all log-transformed variables (in this
case lnca125 & lnbnp)]......dataset imputed [missing values
for log-transformed variables imputed (lnca125_i & lnbnp_i)]

// Back-transformation of imputed variables:

. ds lnca125, d
lnca125         float  %9.0g                  ln(ca125-2.120054)

. gen double ca125_i=exp(lnca125_i)+ 2.120054
. label var ca125_i "exp(lnca125_i)+ 2.120054"
. replace ca125_i=ca125 if ca125~=.

. ds lnbnp, d
lnbnp           float  %9.0g                  ln(bnp-1.681753)

. gen double bnp_i=exp(lnbnp_i)+ 1.681753
. label var bnp_i "exp(lnbnp_i)+ 1.681753"
. replace bnp_i=bnp if bnp~=.

I wouldn't mind to do it for a couple variables. However, when you
have 30-40 continuous variables, you
start looking for ways to speed-up the process.

Any help on how to automate this back-transformation?

Best wishes,

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index