dynamic probit fixed effects model using "probitfe" r(3900)

[nilesh suman]

Hi all,

I am working with large dynamic binary panel data with i=53856 and t=11 (total obs. 592416). My objective is to estimate the dynamic probit (or logit) model using the user-written command "probitfe " in Stata17, please see here https://journals.sagepub.com/doi/abs...867X1701700301

My data looks like the following,

xtdescribe

carid: 21, 33, ..., 370297 n = 36176
year: 2006, 2007, ..., 2017 T = 12
Delta(year) = 1 unit
Span(year) = 12 periods
(carid*year uniquely identifies each observation)

Distribution of T_i: min 5% 25% 50% 75% 95% max
12 12 12 12 12 12 12

Freq. Percent Cum. Pattern

36176 100.00 100.00 111111111111

36176 100.00 XXXXXXXXXXXX


My dependent variable is dynamic binary (0,1) as follows,

Summary for variables: compliance
Group variable: __000000 (Year)

__000000 N Mean SD Min Max

2006 36176 .7172158 .4503586 0 1
2007 36176 .6294781 .4829512 0 1
2008 36176 .570544 .4950054 0 1
2009 36176 .469372 .4990679 0 1
2010 36176 .469372 .4990679 0 1
2011 36176 .4473131 .4972232 0 1
2012 36176 .4584531 .4982778 0 1
2013 36176 .4159664 .4928946 0 1
2014 36176 .3772391 .4847022 0 1
2015 36176 .3432939 .4748152 0 1
2016 36176 .2648717 .441271 0 1
2017 36176 .2130971 .4095013 0 1

Total 434112 .448018 .4972911 0 1



I am getting the following error ;

// Uncorrected estimator: a dynamic model with individual and time effects
logitfe compliance L.compliance $xvar , noc

Computing uncorrected fixed effects estimator

note: multiple positive outcomes within groups encountered
note: multiple positive outcomes within time periods encountered
note: 3235 groups (35585 obs) dropped because of all positive or
all zero outcomes
J(): 3900 unable to allocate real <tmp>[356884,32444]
logit(): - function returned error
<istmt>: - function returned error
r(3900);


I have a 16gb RAM, 1 TB SSD and 9thGenI7 laptop.

I understand that this may be due to a large number of individual effects.

Can anyone suggest to me how to get these estimates running Stata17?

Given my objective is to run dynamic probit fixed effects, I appreciate any alternative suggestions.

Thank you.

[Jared Greathouse]

For anyone to help you, as the FAQ says, you must present an example dataset using dataex as well as the exact code you used in the context of said dataset.

Anyways, Stata says that there are groups with all 1 or 0 outcomes. This means these have to be dropped because the equation perfectly predicted and there's 0 within group variation in that group