egen(cut)

[tim bradshaw]

Hi there

I want to group a continuous variable (var). My code syntax is as follows:

egen vargrp = cut(var), at(10,20,30)

However, I want to cut into groups of 10 way beyond 30 (up to about a thousand). Is there any quick way I can do this without having to list 10,20,30,40,50,60.... etc?

Thanks

[Nick Cox]

This is quite often asked here. Personally I dislike that function as you have to look each time at the code to see exactly what happens at boundaries (the documentation doesn't seem lucid to me) and its conventions for values beyond specified limits are puzzling.

But there are wonderful solutions for what you want. You just need your own generalisations of floor() which rounds down and ceil() which rounds up.

. sysuse auto, clear
(1978 Automobile Data)

. gen length2 = 10 * ceil(length/10)

. gen length3 = 10 * floor(length/10)

. tabdisp length if length2 != length3, c(length?)

----------------------------------
Length    |
(in.)     |    length2     length3
----------+-----------------------
      142 |        150         140
      147 |        150         140
      149 |        150         140
      154 |        160         150
      155 |        160         150
      156 |        160         150
      157 |        160         150
      161 |        170         160
      163 |        170         160
      164 |        170         160
      165 |        170         160
      168 |        170         160
      169 |        170         160
      172 |        180         170
      173 |        180         170
      174 |        180         170
      175 |        180         170
      177 |        180         170
      179 |        180         170
      182 |        190         180
      184 |        190         180
      186 |        190         180
      189 |        190         180
      192 |        200         190
      193 |        200         190
      195 |        200         190
      196 |        200         190
      197 |        200         190
      198 |        200         190
      199 |        200         190
      201 |        210         200
      203 |        210         200
      204 |        210         200
      206 |        210         200
      207 |        210         200
      212 |        220         210
      214 |        220         210
      217 |        220         210
      218 |        220         210
      221 |        230         220
      222 |        230         220
      233 |        240         230
----------------------------------

In short both solutions round, but one returns lower class limits and the other returns upper class limits.

See also http://www.stata-journal.com/sjpdf.h...iclenum=dm0002

I note without approving that many people like int() and round().

[tim bradshaw]

perfect - thank you