Statistics on approval

[starbucks]

Folks,
Talked to a local professor of statistics last night, he is a PhD from
CMU and has gone through the immigration process along with
several of his friends. As per his assessment, the # GCs issued
in a certain time window (say Sep. 2003) by INS, depends on several factors - a) elapsed time since RD of I-485, b) # needed
to be approved based on the state's requirements, c) # needed to be approved based on the GC applicant's country of origin, d)
priority date of the application, e) any special factors
(as he calls it) such as religion/political considerations, non employment etc f) # RFEs issued.

If this is the case, then RD is just one of the factors under consideration. It is quite possible that a person with an RD of
Jan. 2002 in Illinois, could get his/her approval before a person
with an RD of Oct. 2001 from Kansas, because Illinois has a
bigger quota for the month of September 2003. If we were to drill
down further, person with an RD of Jan. 2002 in Illionis, origin -
China, could be approved earlier than a person with an RD of Dec. 2001 in Illinois - origin India.

What this would mean, is any prediction of approval dates for a
population would have to take into account N number of independent variables, and it is possible based on data gathered from approvals so far, a statistical model of approvals could be constructed (however the prof. warned that the variability of the prediction would be very high, due to the data not being clean enough).


starbucks
RD - 12/14/01
ND - 12/15/01
RFE - 6/30/02
RFE response recv'd - 8/7/03
2nd FP - 9/18/03

 

[gheewalabhai888]

we wish ins goes by any so called

"factors". They are a big organization that don't know what they are doing and totally out of control. And no budget to fix it in the future. He may be trying to make sense out of this chaotic statistics, but we on this forum have seen totally out of sync approvals.

 

[d1203]

How did this prof arrive at those assessments ? Did he actually speak to someone from INS or are these just his assumptions based on his personal expriences ?

Did he take into consideration changes in immigration laws post 9/11 ?

 

[Sankrityayan]

Originally posted by d1203
How did this prof arrive at those assessments ? Did he actually speak to someone from INS or are these just his assumptions based on his personal expriences ?

Did he take into consideration changes in immigration laws post 9/11 ?

Once we make the assumption that a data set (like what we have in rupnet) represents a randomly drawn sample (which is a big leap of faith), it is quite simple to figure out which of the available variables exhibit predictive power. You can specify a simple regression model and by trial and error add variables that improve R2 and remove the ones that do not. After a few such iterations, R2 will plateau and you have your solution (within the bounds of the available data, of-course).

To address your second question, it is also easy enough to do this same regression by dividing the dataset into two parts, one covering pre 9/11 and another covering the post 9/11 period. This will isolate the factors that are unique between these two periods and which therefore can be attributed to changes that resulted from 9/11.

Like in all statistics, the caveats are the key, the most critical being that the dataset needs to be a random sample of the population, and should cover an adequately long period of time.

So that boils it down to the dataset being used and rupnet is certainly inadequate. Maybe the professor is supported by the establishment and hence was able to obtain access to a more representative data set, or maybe s/he is just speculating

All said, there is no point in agonizing over INS's fickle methods.