Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: ratio of 2 different surveys estimates? svyset , suest , fpc , nlcom


From   Steven Samuels <[email protected]>
To   [email protected]
Subject   Re: st: ratio of 2 different surveys estimates? svyset , suest , fpc , nlcom
Date   Fri, 21 Jan 2011 17:32:37 -0500



1. Whether two surveys can be combined and how they should be combine two samples depends on the details of each and the purpose of the combination. My allusion to Austin's "person-year" suggestion should alert you to just one of the problems.

2. Even within a single survey, fpcs can differ greatly, so why not in different surveys? Read up on the theory of sampling without replacement and fpc's in any sampling text, e.g. Lohr, Sharon L. 2009. Sampling: Design and Analysis. 2nd ed. Boston, MA: Cengage Brooks/Cole.


3. A google searches "two-frame surveys", "multiple-frame surveys", and "repeated surveys" will turn up many references to combining surveys or samples.

4 . Use fpc's if you want to describe the particular populations; Do not use them use them if you plan analyses of the causal, not descriptive association of two or more factors See point 2 of http://www.stata.com/statalist/archive/2009-02/msg00806.html


Steve
Steven J. Samuels
[email protected]
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783


On Jan 21, 2011, at 12:57 AM, Roman Kasal wrote:

1) in my case the FPCs in my multistage survey are not much different so
not a big problem (but through the time the FPCs can differ - year 2000
and year 2010), but
2) what if I have 2 different multistage surveys that I got (eg found a
survey from Germany on internet and the same survey but with completely
different FPCs). Is really correct to combine FPCs (eg one stratum with
FPC1=0.05 and FPC in the second survey FPC2=0.7)? Could you recommend me
any literature?

thank you!

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Steven
Samuels
Sent: Thursday, January 20, 2011 8:16 PM
To: [email protected]
Subject: Re: st: ratio of 2 different surveys estimates? svyset , suest
, fpc , nlcom

Austin--

I agree that the interpretation of statistics in the total population
is not easy. I recall that you once suggested that the unit for
repeated surveys of the same population should be something like a
person-year.  However Roman seems to be interested in estimating year-
to-year differences (or ratios), so, for his purposes, the definition
of the total population might not be an issue.


The -svyset- statement that I wrote assumes that, aside from certainty
PSUs, a new sample of PSUs was drawn in each year. If this is not
correct, then a different -svyset- statement is needed.

Roman, I'd like to know why fpc's varied greatly and how they were
computed.  If the survey had several stages and you are interested in
primarily descriptive statistics, then perhaps the -svyset- should
reflect the multi-stage structure.


Steve



On Jan 20, 2011, at 12:43 PM, Austin Nichols wrote:

Steve--
This works for inference on a population that is the sum of the
populations in each of two years--a strange population for many
purposes.

On Thu, Jan 20, 2011 at 9:00 AM, Steven Samuels <[email protected]>
wrote:

"Is this statistically correct..?"
Yes

Steve
[email protected]


On Jan 20, 2011, at 4:03 AM, Roman Kasal wrote:

ok, I understand that, I already did this once, but is this
statistically correct adjusting=combine FPCs of two different surveys?
What if the FPCs are really very different (this is not mentioned
case)?

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Steven
Samuels
Sent: Wednesday, January 19, 2011 9:08 PM
To: [email protected]
Subject: Re: st: ratio of 2 different surveys estimates? svyset ,
suest
, fpc , nlcom

Another way of putting this: When you combine independent surveys, you
must create a new stratum variable in which the individual original
strata in each survey appear. You would do this in your data, for
example by:
****************************
egen stratum =group(year stk)
*****************************
If there were 10 strata in 2006 and 10 in 2007, the new stratum
variable should have 20 levels. Then a single -svyset- statement will
allow joint analysis of the two surveys, as in my example.

Steve
[email protected]
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index