Checking for Hetro in Longitudinal (Panel) Data

[Shifa Fatimah]

Good Day,
I am trying to check for hetro issue in my longitudinal data where the N is large and P is small. I have run the following command:

xtgls mig_dummy lagSPEI i.age_cat i.Gender , igls panels(heteroskedastic)
estimates store hetero
xtgls mig_dummy lagSPEI i.age_cat i.Gender
local df = e(N_g) - 1
latest hetero1 . , df(`df'

but I get the error
matsize too small - should be at least 2157
r(908);

I have also aplied the command
set matsize 200, permanently
But the error is still percisting.

My data from the dataex command is:

[CODE]
* Example generated by -dataex-. To install: ssc install dataex
clear
input float(mig_dummy lagSPEI age_cat) byte Gender
0 -.376669 30 2
0 .751355 30 2
0 1.18046 30 2
0 .785417 30 2
0 1.81745 30 2
0 1.01097 30 2
0 1.50308 30 2
0 .0247821 30 2
0 -.41504 30 2
0 -.689454 30 2
0 1.34077 30 2
0 .965805 30 2
0 1.14084 30 2
0 1.51453 30 2
1 .354391 30 2
0 .751355 30 1
0 1.18046 30 1
0 .785417 30 1
0 1.81745 30 1
0 1.01097 30 1
0 1.50308 30 1
0 .0247821 30 1
0 -.41504 30 1
0 -.689454 30 1

Could anyone please advise me how to proceed with it as I need to check at least for hetro.
Thanking you

[Carlo Lazzaro]

Shifa:
if you have a N>T panel dataset, you should consider -xtreg-.

[Shifa Fatimah]

Thank you for your response. I have been over your discussuion about xtreg on

https://www.statalist.org/forums/for...roskedasticity

I wanted to ask should I do xtreg on categorical variables as well, my dependent variable is a dummy.

[Carlo Lazzaro]

Shifa:
you should switch to -xtlogit- or -xtprobit-. However, please note that -xtlogit- allows conditional fixed effect, whereas -xtprobit- does not allow -fe- at all.