What is the Test for Under-Dispersion in STATA?

[Shashank Shukla]

We are trying to fit a zero-weighted binomial model to our data set. The model is strongly supported by theory. In that context, we shall be grateful for any help on some of our problems:
(1) Like the way one can formally test for overdispersion in count data, can one test for underdispersion as well? If yes, what test and how to do that in STATA?
2) We would like to test whether zero-weighted binomial regression model is indeed a better model for our data set than the simple binomial regression model. Both AICs and BICs suggest that the zero-weighted model is the better one. However, can one test in STATA whether all the "inflation" parameters are zero jointly? If that test rejects the null hypothesis, can one then also comment that the zero-weighted model is a necessity here?

[Shashank Shukla]

A small correction. The model which we have is Zero inflated binomial regression. We need guidance on the following:
1) Is there a test for under-dispersion in STATA
2) Can one test in STATA whether all the "inflation" parameters are zero jointly? If that test rejects the null hypothesis, can one then also comment that the zero-weighted model is a necessity here?

[Joseph Coveney]

A small correction. The model which we have is Zero inflated binomial regression. . . .
1) Is there a test for under-dispersion in STATA

I don't know what a zero-inflated binomial model is, but if you're interested in examing underdispersion in a binomial model, then maybe you can look at the upper confidence bound in something like the following.

.ÿ
.ÿversionÿ16.0

.ÿ
.ÿclearÿ*

.ÿ
.ÿsetÿseedÿ`=strreverse("1524638")'

.ÿ
.ÿmatrixÿdefineÿMÿ=ÿJ(1,ÿ5,ÿ0.5)

.ÿmatrixÿdefineÿCÿ=ÿJ(5,ÿ5,ÿ-0.15)

.ÿforvaluesÿiÿ=ÿ1/2ÿ{
ÿÿ2.ÿÿÿÿÿmatrixÿdefineÿC[`i',ÿ`i']ÿ=ÿ1
ÿÿ3.ÿ}

.ÿovbdÿ,ÿstub(rsp)ÿmeans(M)ÿcorr(C)ÿn(250)ÿclear
(obsÿ250)

.ÿ
.ÿgenerateÿintÿpidÿ=ÿ_n

.ÿquietlyÿreshapeÿlongÿrsp,ÿi(pid)ÿj(tim)

.ÿ
.ÿ*
.ÿ*ÿBeginÿhere
.ÿ*
.ÿprogramÿdefineÿbootem,ÿrclass
ÿÿ1.ÿÿÿÿÿÿÿÿÿversionÿ16.0
ÿÿ2.ÿÿÿÿÿÿÿÿÿsyntax
ÿÿ3.ÿ
.ÿÿÿÿÿÿÿÿÿxtgeeÿrspÿi.tim,ÿi(newpid)ÿfamily(binomial)ÿlink(logit)ÿcorr(exchangeable)
ÿÿ4.ÿÿÿÿÿÿÿÿÿestatÿwcorrelation
ÿÿ5.ÿÿÿÿÿÿÿÿÿtempnameÿRÿrho
ÿÿ6.ÿÿÿÿÿÿÿÿÿmatrixÿdefineÿ`R'ÿ=ÿr(R)
ÿÿ7.ÿÿÿÿÿÿÿÿÿscalarÿdefineÿ`rho'ÿ=ÿ`R'[1,ÿ2]
ÿÿ8.ÿÿÿÿÿÿÿÿÿreturnÿscalarÿrhoÿ=ÿ`rho'
ÿÿ9.ÿend

.ÿ
.ÿbootstrapÿrhoÿ=ÿr(rho),ÿreps(50)ÿcluster(pid)ÿidcluster(newpid):ÿbootem
(runningÿbootemÿonÿestimationÿsample)

Bootstrapÿreplicationsÿ(50)
----+---ÿ1ÿ---+---ÿ2ÿ---+---ÿ3ÿ---+---ÿ4ÿ---+---ÿ5ÿ
..................................................ÿÿÿÿ50

BootstrapÿresultsÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿÿÿÿÿ=ÿÿÿÿÿÿ1,250
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿReplicationsÿÿÿÿÿÿ=ÿÿÿÿÿÿÿÿÿ50

ÿÿÿÿÿÿcommand:ÿÿbootem
ÿÿÿÿÿÿÿÿÿÿrho:ÿÿr(rho)

ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ(Replicationsÿbasedÿonÿ250ÿclustersÿinÿpid)
------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿObservedÿÿÿBootstrapÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNormal-based
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿÿCoef.ÿÿÿStd.ÿErr.ÿÿÿÿÿÿzÿÿÿÿP>|z|ÿÿÿÿÿ[95%ÿConf.ÿInterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿrhoÿ|ÿÿ-.1423329ÿÿÿ.0086542ÿÿÿ-16.45ÿÿÿ0.000ÿÿÿÿ-.1592948ÿÿÿÿ-.125371
------------------------------------------------------------------------------

.ÿestatÿbootstrap,ÿall

BootstrapÿresultsÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿÿÿÿÿ=ÿÿÿÿÿÿ1,250
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿReplicationsÿÿÿÿÿÿ=ÿÿÿÿÿÿÿÿÿ50

ÿÿÿÿÿÿcommand:ÿÿbootem
ÿÿÿÿÿÿÿÿÿÿrho:ÿÿr(rho)

ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ(Replicationsÿbasedÿonÿ250ÿclustersÿinÿpid)
------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿObservedÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿBootstrap
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿÿÿCoef.ÿÿÿÿÿÿÿBiasÿÿÿÿStd.ÿErr.ÿÿ[95%ÿConf.ÿInterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿrhoÿ|ÿÿ-.14233287ÿÿÿ.0000755ÿÿÿÿ.0086542ÿÿÿ-.1592948ÿÿÿ-.125371ÿÿÿ(N)
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ-.158442ÿÿ-.1258217ÿÿÿ(P)
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ-.158442ÿÿ-.1258217ÿÿ(BC)
------------------------------------------------------------------------------
(N)ÿÿÿÿnormalÿconfidenceÿinterval
(P)ÿÿÿÿpercentileÿconfidenceÿinterval
(BC)ÿÿÿbias-correctedÿconfidenceÿinterval

.ÿ
.ÿexit

endÿofÿdo-file


.

The user-written command -ovbd- is available from SSC. Note that it requires the user-written command -ridder-, too.

[Marcos Almeida]

You may wish to read this thread.

[Shashank Shukla]

Many thanks Marcos for sharing the thread.