5:07 p.m.
Hi
Is there any option to create grand mean centering of a varibale x_mc, (x_i - x_bar),
group mean (x_barj), centering within clustered variable x_cwc (x_i - x_barj). These are
necesary functions to teach mixed models. If there are not, i request you to provide.
7:19 p.m.
Am 06.04.2026 um 19:07 schrieb Ramki S:
Hi
Is there any option to create grand mean centering of a varibale x_mc, (x_i - x_bar),
group mean (x_barj), centering within clustered variable x_cwc (x_i - x_barj). These are
necesary functions to teach mixed models. If there are not, i request you to provide.
- general centering: Add/Standardize, then choose only centering; script
function "stdize"
- Demeaning by a discrete group indicator (possibly more general than a
panel group): I'm not aware of a one-stop-shopping in the GUI here, but
one could do: Add/Dummies for discrete variables..., Model/OLS..., then
add those created dummies as regressors, estimate that model, and from
the model window save the residuals as series.
(Possibly the stdize function could be extended to allow a factorizing
variable. There also a view of factorized basic statistics like the
mean, but there I don't think it's possible to save those means to the
dataset.)
For scripting you could do the same things, using stuff like the
dummify() function, and/or aggregate() etc.
-s
9:06 p.m.
On 06/04/2026 19:07, Ramki S wrote:
Hi
Is there any option to create grand mean centering of a varibale x_mc, (x_i - x_bar),
group mean (x_barj), centering within clustered variable x_cwc (x_i - x_barj). These are
necesary functions to teach mixed models. If there are not, i request you to provide.
The aggregate() function is probably your best friend here:
<hansl>
open mroz87.gdt
matrix m = aggregate(WA, KL6, mean)
series grpmeans = replace(KL6, m[,1], m[,3])
series WA_demeaned = WA - grpmeans
boxplot WAdemeaned KL6 --fact --output=display
</hansl>
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
4:14 a.m.
Aggregate is a nice function. Can we make it gui based.
function: cluster_mean(y, unit){
matrix m = aggregate(y, unit, mean)
series y_cm = replace(unit, m[,1], m[,3])
}
function: centered_within_cluster(y, unit){
matrix m = aggregate(y, unit, mean)
series cluster_mean = replace(unit, m[,1], m[,3])
series y_cwc = y - cluster_mean
}
5:22 a.m.
Hi,
I tried my first package to calculate grand mean centered, cluster mean and centered
within cluster for a given variable. I made some mistake some where that the menu is not
working properly. Can anyone suggest how to proceed from there. Also how to edit a GFN
file once it is created.
https://www.dropbox.com/scl/fi/xbubchxgfgwwe4quf1we0/ClusterMeans.gfn?rlk...
8:57 a.m.
Seems I fixed the issue. However I want to add the suffix to _gmc, _cwc, _cm to grand mean
centered, cluster mean centered, cluster mean variables automatically. How to do this?
Currently i have to give a variable name.
Without providing a name if i run the script, a new window with ? grand_mean_centered(y)
is opening.
9:10 a.m.
Am 07.04.2026 um 10:57 schrieb Ramki S:
Seems I fixed the issue. However I want to add the suffix to _gmc,
_cwc, _cm to grand mean centered, cluster mean centered, cluster mean variables
automatically. How to do this? Currently i have to give a variable name.
Either string substitution with the @ symbol or the genseries() function.
string yname = "GPA"
series @yname_cm = cluster_mean(@yname, PSI)
or
genseries(yname ~ "_cm", cluster_mean(GPA, PSI))
For the latter line, if you're inside a function with input series
argument y, something like:
genseries(argname(y) ~ "_cm", cluster_mean(y, PSI))
We tend to discourage string substitution unless really necessary, but
it's official syntax and sometimes more intuitive.
Without providing a name if i run the script, a new window with ?
grand_mean_centered(y) is opening.
Sorry, I don't understand what you did here.
-s
14
days inactive
15
days old
8 comments
3 participants
participants (3)
-
Ramki S -
Riccardo (Jack) Lucchetti -
Sven Schreiber