# st: Dispersion of workplace years of training

 From peter harper To statalist@hsphsun2.harvard.edu Subject st: Dispersion of workplace years of training Date Tue, 6 Jun 2006 11:39:24 +0000 (GMT)

```Dear Statalist

Thanks Austin and Nick. I am now looking at the dispersion measure of average of absolute differences between own training (across random sample of workers) and mean workplace
training.

I am working on a cross-section dataset, with different firms and
different individuals. How would one generate a variable which would
provide a dispersion of workplace years of training: absolute mean diff.
across workers based on Ti. The dispersion measure is the average of
absolute differences between own training, tij (across the random sample
of workers) and mean workplace training, Tj.

Sigma (this is the greek letter for sum)|(Tj - tij)|/nj

where tij = workers years of training, i = worker and j = firm and Tj is
given below.
I already have a variable for mean workplace years of training called Tj:
which is based on percentage of the workforce in each of the k occupations
times average years of training for that occupation from worker
respondents.
Would the command be:

where meantrwk=meanworkplace years of training; b4a2=workers who have been trained.

But when I run the regression with both mean dispersion of education and mean dispersion of training. The mean dispersion of training is dropped. I then find the mean is zero. So, I would be grateful if anyone would let me know, the correct command for the mean dispersion of training as mentioned above.

Thanks

Nick Cox wrote:
The mean (absolute) deviation (from the mean)
is also coded officially as -egen-'s -mdev()-
function.
The mean deviation is, NB, not much robust (resistant)
than the standard deviation, as it is mean-based.
So, if you are playing with the idea of robustness,
do not stop there:
more robust yet is the median (absolute)
deviation (from the median), coded officially
Nick
n.j.cox@durham.ac.uk

Austin Nichols
> peter harper:
> One measure of dispersion of years of education by workplace would be
> . egen sde=sd(educ), by(workplace)
> but if you want the mean absolute difference (AD), you can code:
> . egen ej=mean(educ), by(workplace)
> for example.
> On 5/15/06, peter harper <pharper11@yahoo.co.uk> wrote:
> > I am working on a cross-section dataset, with different
> firms and different individuals. How would one generate a
> variable which would provide a dispersion of workplace years
> of education: absolute mean diff. across workers based on
> mean worker years of education, say, Ei.
> >
> > I already have a variable for mean workplace years of
> education called Ej: which is based on percentage of the
> workforce in each of the k occupations times average years of
> education for that occupation from worker respondents.
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```