Credibility Intervals
Locked
 
|
Hello,
I am interested in calculating credibility intervals for advertising research. Is possible to calculate credibility intervals in SPSS Stats 23? Best, Bryan |
Administrator
|
Nabble is showing that your message has not (yet) been delivered to the actual mailing list. That may be why there have been no replies yet.
I'm not an expert on credible intervals, but a quick Google search turned up some notes by John Uebersax: http://www.john-uebersax.com/stat312/17%20Credible%20Intervals%20and%20Confidence%20Intervals.pdf At the bottom of p. 2, he says: "The real reason people continue to use confidence intervals, ironically, is that, under many conditions, they are identical to credible intervals. That is, people for decades have simply constructed confidence intervals, and then interpreted them as credible intervals." If credible intervals are identical to confidence intervals under many conditions, I think you need to provide more information. Credible intervals for what, exactly? HTH.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
|
|
Hi Bruce,
Thanks for that pdf. Credible intervals for a beta distribution. Best, Bryan |
|
Administrator
|
This post was updated on .
Let me stress again that I am not an expert on Bayesian statistics. But after reading the notes linked below (1st link), I wonder if you are asking how to compute Jeffreys prior intervals for binomial proportions. If so, see the second link and the code from one of my old syntax files (pasted below).
http://ocw.jhsph.edu/courses/MethodsInBiostatisticsI/PDFs/lecture13.pdf http://www-01.ibm.com/support/docview.wss?uid=swg21474963 * Jeffreys prior interval for proportion = x/n. COMPUTE lower = idf.beta(alpha/2,x+.5,n-x+.5). COMPUTE upper = idf.beta(1-alpha/2,x+.5,n-x+.5). HTH. EDIT: Here is another document that provides support for the SPSS computations shown above. https://support.sas.com/resources/papers/proceedings15/3020-2015.pdf
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
|
|
Like Bruce, I am not an expert on credible intervals or Bayesian
statistics but I believe that credible intervals are calculated on posterior distributions which would make sense since you want to have an interval for an estimate from the analysis one would do. A number of program other than SPSS can calculate credible intervals for beta and other distributions (e.g. WinBugs, Matlab, etc.). R can calculate credible intervals for beta distributions and you can see how it is done on the following website but I warn you that it is some fugly code: https://www.r-bloggers.com/the-beta-prior-likelihood-and-posterior/ The good news, I think, is that if you can access R through your SPSS installation, you may be able to apply this program to your data. If you don't have R installed, then there is another option -- which should make Jon Peck snicker at the rest of us -- through python. Jon may be able to provide more advice on how to do this through SPSS; see: http://freakonometrics.hypotheses.org/18117 Then again, you might be able to do the necessary analysis in Excel. Here is a link to the book by Meeker et al "Statistical Intervals" (2nd Ed.) website which has a section for add-in software for Excel to do the analyses that they cover in their book (a PDF of the front matter of the book, that is, the table of contents and the prefaces is available on web so you can google it -- in their preface they mention that there are variety of programs that can be used for their analyses but since can become outdated, they recommend the use of their software which they supposedly will keep updated); see the following for the first webpage introducing the book: http://bcs.wiley.com/he-bcs/Books?action=index&bcsId=10767&itemId=0471687170http://bcs.wiley.com/he-bcs/Books?action=index&bcsId=10767&itemId=0471687170 You can access the software by clicking on "Browse by Resources" tab near the top or click on the following: http://bcs.wiley.com/he-bcs/Books?action=resource&bcsId=10767&itemId=0471687170&resourceId=42630 HTH. -Mike Palij New York University [hidden email] ----- Original Message ----- On Sunday, June 11, 2017 9:06 AM, Bruce Weaver wrote: > Let me stress again that I am not an expert on Bayesian statistics. > But > after reading the notes linked below (1st link), I wonder if you are > asking > how to compute Jeffreys prior intervals for binomial proportions. If > so, > see the second link and the code from one of my old syntax files > (pasted > below). > > http://ocw.jhsph.edu/courses/MethodsInBiostatisticsI/PDFs/lecture13.pdf > http://www-01.ibm.com/support/docview.wss?uid=swg21474963 > > * Jeffreys prior interval for proportion = x/n. > COMPUTE lower = idf.beta(alpha/2,x+.5,n-x+.5). > COMPUTE upper = idf.beta(1-alpha/2,x+.5,n-x+.5). > > HTH. > > Bryan Mac wrote >> Hi Bruce, >> >> Thanks for that pdf. Credible intervals for a beta distribution. >> ===================== 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 |
|
|
In reply to this post by Bryan Mac
I need to correct the info I provided below about the
Meeker et al webpage. They do not provide an Excel add-in but do provide an R program for the intervals they cover in their book. Quoting from the Preface: |The webpage for this book, created by Wiley, can |be found at www.wiley.com/go/meeker/intervals. |In addition to the link |to the StatInt R package and |the Excel spreadsheet on current statistical interval |capabilities of popular software, this webpage provides |some tables and figures from the first edition that are |omitted in the current edition, as well as some |additional figures and tables, for finding statistical intervals. | |We plan to update this webpage periodically by adding new |materials and references, (numerous, we hope) reader comments |and experiences, and (few, we hope) corrections. Sorry about that. I shouldn't post before my first cup of coffee. ;-) -Mike Palij New York University [hidden email] ----- Original Message ----- On Sunday, June 11, 2017 9:34 AM, Mike Palij wrote: [snip] > Then again, you might be able to do the necessary analysis > in Excel. Here is a link to the book by Meeker et al > "Statistical Intervals" (2nd Ed.) website which has a section > for add-in software for Excel to do the analyses that they > cover in their book (a PDF of the front matter of the book, > that is, the table of contents and the prefaces is available > on web so you can google it -- in their preface they mention > that there are variety of programs that can be used for their > analyses but since can become outdated, they recommend > the use of their software which they supposedly will keep > updated); see the following for the first webpage introducing > the book: > http://bcs.wiley.com/he-bcs/Books?action=index&bcsId=10767&itemId=0471687170http://bcs.wiley.com/he-bcs/Books?action=index&bcsId=10767&itemId=0471687170 > > You can access the software by clicking on "Browse by Resources" > tab near the top or click on the following: > http://bcs.wiley.com/he-bcs/Books?action=resource&bcsId=10767&itemId=0471687170&resourceId=42630 ===================== 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 |
|
|
The StatInt package is not currently available on CRAN and requires compilation of Fortran code, so I expect this would be a pain for most folks to install. The description on the cited page says that it has been submitted to CRAN, but that text has no date on it. The Browse by chapter menu is empty, however the browse by resource dropdown does come up with an Excel spreadsheet. The spreadsheet, though, doesn't seem to offer any calculation code. On Sun, Jun 11, 2017 at 8:28 AM, Mike Palij <[hidden email]> wrote: I need to correct the info I provided below about the |
|
|
Hi Jon,
My experience with publishers' websites for
texts has been very
mixed and I am not surprised that this one
disappoints. Sometimes
it is helpful to go to the senior author's
website to see if there is
more information available there. In
this case, I went to William
Meeker's website at Iowa State University and
was surprised
to find nothing on the second edition but
there was info for the
1991 edition (Hahn & Meeker);
see:
There is even a page for a DOS-based program
(named StInt)
to do the analyses in the text; see:
I find this situation very odd which suggests
to me that perhaps
Meeker did not have very much to do with the
2nd edition
and perhaps the third author (Escobar) did
most of the
updating (this kind of situation was done with
Fred Kerlinger's
"Foundations of Behavioral Research" 2000
edition but
the second author used some research examples
that are
now considered to be either severely mistaken
or just
plain fraud). Anyway, it is unclear who
is responsible for
the R version and getting it accepted by
CRAN.
On another note, I tried to correct my
comments about
the Excel add-in with my quote from the
preface of the
2nd edition which says that it just contains
references,
tables and figures.
Jon, any comments on this link which I
provided in my
earlier post:
or this one:
-Mike Palij
New York University
===================== 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 |
|
|
I'm deep in something else right now but will look at those links when I come up for air. I might remind folks who are interested that there are four Bayesian extension commands for Statistics - STATS_BAYES_ANOVA, STATS_BAYES_CROSSTABS STATS_BAYES_REGR, and STATS_BAYES_TTEST although none bears directly on the Beta distribution issue. As for publishers' websites, I better go look at ours, which is supposed to have the data and code for the book. :-) On Sun, Jun 11, 2017 at 3:21 PM, Mike Palij <[hidden email]> wrote:
|
|
Administrator
|
In reply to this post by Bryan Mac
Bryan, it would be nice to get some closure on this thread. Have any of the responses provided what you were looking for? Thanks.
p.s. - The entire thread can be viewed here: http://spssx-discussion.1045642.n5.nabble.com/Credibility-Intervals-td5734406.html.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
|
|
Hi everyone,
Thank you for all the great responses! I have read most of the links provided and I found something that I can use. Jon, if you have some idea on how to integrate the credible intervals in R Essentials for SPSS that would be great as well. Best, Bryan |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |