NCES data and SPSS R plugin

[Lynette Osborne]

Hi Everyone,

I'm going to the be doing some work with NCES data soon and I've been struggling with how to do balanced repeated replication (BRR) estimation or other replication based methods of variance estimation, (I'd prefer jackknifing or bootstrapping methods but those aren't available apparently) since it looks like Taylor Series is my only option.  I spoke with the SPSS folks and they said that an R plugin would be my only option.  Has anyone used the R plugin for complex data like the NCES?  And, if so, do you have any suggestions or tips for someone new to it?

Thanks so much!

Lynette

[David Marso]

You can always roll your own jackknife and/or bootstrap.  Search these archives for ideas!
Other than that maybe you should reference the link to whatever R plugin you are talking about.
--

Lynette Osborne wrote

Hi Everyone,
I'm going to the be doing some work with NCES data soon and I've been
struggling with how to do balanced repeated replication (BRR) estimation or
other replication based methods of variance estimation, (I'd prefer
jackknifing or bootstrapping methods but those aren't available apparently)
since it looks like Taylor Series is my only option.  I spoke with the SPSS
folks and they said that an R plugin would be my only option.  Has anyone
used the R plugin for complex data like the NCES?  And, if so, do you have
any suggestions or tips for someone new to it?
Thanks so much!
Lynette

[David B. Nolle-2]

Lynette and David (Marso),

I recommend the use of WesVar 5.1 for Replication-Based Variance Estimation
for Analysis of Complex Survey Data.  WesVar is freely available from Westat
where outstanding statisticians have created an excellent user-friendly
program which uses SPSS and other formats for input. I suggest that you look
at
http://www.westat.com/Westat/expertise/information_systems/WesVar/index.cfm
for details.

I personally have used WesVar for many years for national face-to-face
surveys as well as telephone surveys to handle the effects of
stratification, clustering, and weighting in complex survey designs. I think
that the manual is well-documented. In the past, I have also used the
Complex Samples module within SPSS and have  found that its Taylor expansion
estimates are very close to my jackknifed estimates from WesVar. I have also
used Bob Fay's CPLEX in the very distant past and have found that the
jackknifed estimates from CPLEX are virtually identical to jackknifed
estimates from WesVar.  Thus, I think that WesVar has withstood the tests of
time.

I apologize for the delay in my response because I use a daily digest to
summarize my messages.

David B. Nolle

Date:    Mon, 23 Dec 2013 05:33:35 -0800
From:    David Marso <[hidden email]>
Subject: Re: NCES data and SPSS R plugin

You can always roll your own jackknife and/or bootstrap.  Search these
archives for ideas!
Other than that maybe you should reference the link to whatever R plugin you
are talking about.
--

Lynette Osborne wrote

> Hi Everyone,
> I'm going to the be doing some work with NCES data soon and I've been
> struggling with how to do balanced repeated replication (BRR)
> estimation or other replication based methods of variance estimation,
> (I'd prefer jackknifing or bootstrapping methods but those aren't
> available
> apparently)
> since it looks like Taylor Series is my only option.  I spoke with the
> SPSS folks and they said that an R plugin would be my only option.
> Has anyone used the R plugin for complex data like the NCES?  And, if
> so, do you have any suggestions or tips for someone new to it?
> Thanks so much!
> Lynette

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD