st: RE: Matching Problem

[Nick Cox <n.j.cox@durham.ac.uk>]

I support Stefan's approach here. Note that (e.g.) 

gen match = 0 
replace match = 1 if int(test/5) == int(train/5)

is just 

gen match = int(test/5) == int(train/5) 

However, there is a small problem with either if there are missing values. 

gen match = int(test/5) == int(train/5) if !missing(test, train) 

is better than either. 

Nick 
n.j.cox@durham.ac.uk 

Stefan.Gawrich@hlpug.hessen.de

try this:

*** Example Dataset
clear
set obs 1000
gen test = int(uniform()*200)
gen train = int(uniform()*200)
gen match = 0

*** You're not clear about start and end of the intervals you want: 1, 6, 11 vs. 70, 125
*** So choose the command that fits.

** Intervals [0..4] [5..9] [10..14]... [195-199] [200]
replace match = 1 if int(test/5) == int(train/5) 

** Intervals [0] [1..5] [6..10] [11..15]... [196..200]
replace match = 1 if int((test+4)/5) == int((train+4)/5)

Raphael Fraser 

I have two variables which I will call test and training each
containing integers between 1-200. Each number in the test variable
represents an image. The training variable contains the closest image
to the test image in terms of similarity.

test = (129, 163, 71, 176, 125, ...)
train = (128, 162, 71, 119, 123, ...)

The objective is to match both integers to the same interval. These
are the intervals 1-5, 6-10,11-15, ..., 196-200. For example, when
test=129 and train=128 are both in the same interval 125-130. Also
test=71,train=71 are both in the interval 70-75. These are successful
mappings. I would like a successful mapping to be =1 and failure=0.


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/