New extension command for SPSS Statistics - STATS POWER
Locked
 
|
A new extension command, STATS POWER, that does power calculations for common statistical tests has been posted. This procedure calculates power and related statistics for a variety of tests. The tests include t tests, proportion tests, anova, correlation, chi-squared, and linear models. Each type of test takes a set of input parameters that vary according to the test. The resulting power can be computed or the reverse. That is, the power can be specified and one of the other parameters omitted, and the procedure will calculate the required value of the input to achieve the specified power. There are two versions of the extension command bundle. STATS POWER is the full procedure, including a dialog box interface. It requires Statistics V24 with FixPack 1 or later. STATS POWER23, for versions 23 and earlier, is the same except that it does not include the dialog box. There are two versions, because the dialog box uses new features in the V24 Custom Dialog Box builder. Both versions require the R Essentials. These commands can be installed from the Utilities menu in V22-23 or the Extensions menu in V24 or obtained from the SPSS Community website. |
Re: New extension command for SPSS Statistics - STATS POWER
Administrator
|
Thanks Jon. I have a few questions.
1. How does STATS POWER differ from SamplePower (https://www-01.ibm.com/marketing/iwm/iwmdocs/tnd/data/web/en_US/trialprograms/U741655I36057W80.html)? (I've never had access to it with my SPSS license, so I don't know much about it.) 2. I'm curious about why R was required. What particular functionality is missing from native SPSS (or Python)? 3. Does STATS POWER include simulation for more complex designs where there may be no straightforward simple equations? Cheers, Bruce
--
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/). |
Re: New extension command for SPSS Statistics - STATS POWER
|
|
Bruce,
SamplePower is a re-branding of Michael Borenstein's Power and Precision. If STATS POWER relies on R, check to see what R functionality and packages it uses. -- Tony Babinec -- ASA Council of Chapters Chair, -- Joint Statistical Meetings 2017 Program Committee -- [hidden email] -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver Sent: Friday, September 23, 2016 9:20 AM To: [hidden email] Subject: Re: New extension command for SPSS Statistics - STATS POWER Thanks Jon. I have a few questions. 1. How does STATS POWER differ from SamplePower (https://www-01.ibm.com/marketing/iwm/iwmdocs/tnd/data/web/en_US/trialprogra ms/U741655I36057W80.html)? (I've never had access to it with my SPSS license, so I don't know much about it.) 2. I'm curious about why R was required. What particular functionality is missing from native SPSS (or Python)? 3. Does STATS POWER include simulation for more complex designs where there may be no straightforward simple equations? Cheers, Bruce ===================== 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 Bruce Weaver
You might not have heard that the IBM SPSS SamplePower program is no longer available from IBM. It was an extra cost option. STATS POWER or STATS POWER23 provides most of the same basic functionality, but it does not have the tutorial aspects of SamplePower. However, it is free. The extension could have been implemented from scratch without R, but I used the R pwr module for the core functionality rather than reimplementing the formulas. Both programs only provide basic functionality. There are much more elaborate but expensive programs available that go far beyond either of these such as PASS as well as some other free alternatives. On a related note, I should mention the STATS OPTIMAL DESIGNMC extension command. Here's a portion of the dialog help. Generate Optimal Experimental DesignThis procedure generates an optimal experimental design dataset using Monte Carlo methods and the Federov algorithm. The procedure can handle an unlimited number of scale (continuous) and categorical variables, but due to layout constraints, the dialog box allows for only two variables. You can start by using the dialog box and then paste and edit the syntax to extend the variable specifications.x On Fri, Sep 23, 2016 at 8:20 AM, Bruce Weaver <[hidden email]> wrote: Thanks Jon. I have a few questions. |
|
|
Jon, thanks for clearing up the story about
the availability of SamplePower
though I am a long time users of PASS
(which invariably comes with NCSS,
a general purpose stat analysis package, or at
least it used -- is it wrong for
me to say I'm still using PASS
2000?).
To be clear, however, the extension commands
depend upon the R power
analysis modules to do the heavy lifting,
right? I think some SPSS users
would prefer a power analysis/sample size
determination procedure built
into SPSS (I rarely use SAS but doesn't it
have these as part of their
native procedures?). I say this because
it seems that instead of building
in capabilities into SPSS proper, R and python
have to be called and
one has to pray to the random number gods that
these programs have
been correctly installed and are not affected
by other factors (e.g.,
oddities in the operating system, phase of the
moon errors, etc.).
I'm still using SPSS ver 23 (the university is
being cheap about upgrading)
and I'm dumb enough to be running one copy on
an XP box (which
is a little like having your finger nails
pulled out with pliers), so getting
everything to play nice is something of a task
(I avoid use the extension
commands for this reason).
But this is useful to know for when I will
upgrade my hardware and software
next year.
-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 |
|
|
Once an extension is installed, it would be difficult for a user to tell that it is not native, although extensions give the secret away by the use of a special icon in the menus.
Installation of the extension frameworks for Python and R has gotten easier and easier with Statistics releases. Python is now part of the main install. R is still separate, but it is usually pretty easy. And it's a one-time task. The benefits, of course, go far beyond any one extension command, since there are well over 100 available for Statistics. Installing one that is not included in the Essentials is easy to do from the menus since V22, and V24 has greatly improved the interface for this. Extensions also open up Statistics for users to implement their own procedures: it puts them in control. It also allows developers to be more productive and to tap the best of what's out there, to the benefit of all users even if they don't have or don't want to use Python, R, or Java skills. That's a win for everyone. More features and faster. Resources are always limited. So, in my opinion, what's left comes down mostly to trust. That's an issue for all open source software, but it is present for traditional commercial software, too, although commercial software likely has more process guarantees. Extensions do, though, come with all the source code, so they are open to inspection and correction by users with the requisite skills. In my substantial experience with R packages from CRAN, I have often found bugs, including sometimes wrong results. More often the bugs result in obvious failure or are due to poor documentation. So I think developers of extensions, whether from IBM or elsewhere, have an obligation to verify the results as much as possible by various means, but this can rarely be definitive. When I have found a bug in a package, I have either worked around it, gotten the author to correct the package, or abandoned it. (Of course, there are never bugs in my code, but I always fix them promptly.) There is always room for improvement, but I think SPSS users are substantially better off because of this technology. On Friday, September 23, 2016, Mike Palij <[hidden email]> wrote:
-- ===================== 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 |
Re: New extension command for SPSS Statistics - STATS POWER
|
Administrator
|
In reply to this post by Mike
Yes, thanks for clarifying, Jon.
Adding to Mike's reasons for preferring functionality being built into SPSS proper, Python and R may not be included when SPSS is installed in student labs on campus. (I just checked yesterday at my own university, and was told that Python was installed, but not R.) By the way, I too have had PASS for several years, and quite like it. Stata also has fairly extensive capabilities built right in (http://www.stata.com/features/power-and-sample-size/). I haven't used it as much as PASS, but when I have, it has been very satisfactory too. Cheers, Bruce
--
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/). |
|
|
"Python and R may not be included when SPSS is installed in student labs on campus. (I just checked yesterday at my own university, and was toldthat Python was installed, but not R.)" Isn't that backwards? The faculty should be telling the IT folks what they need, not the other way around. On Fri, Sep 23, 2016 at 12:00 PM, Bruce Weaver <[hidden email]> wrote: Yes, thanks for clarifying, Jon. |
Re: New extension command for SPSS Statistics - STATS POWER
|
Administrator
|
Yes...in an ideal world. ;-)
--
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/). |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |