Meta-Analysis Resources

Tools for Those Who Summarize the Evidence Base

Resources and networking for those who conduct or interpret meta-analyses related to any phenomenon that is gauged in multiple studies.

Although technically out of print, DSTAT 1.x exists in two main versions:

  • Johnson, B. T. (1993). DSTAT 1.10: Software for the meta-analytic review of research literatures. Mahwah, NJ: Erlbaum.(Version 1.10; software and manual)
  • Johnson, B. T. (1989). DSTAT: Software for the meta-analytic review of research literatures. Mahwah, NJ: Erlbaum.(Version 1.00; software and manual)

The 1993 version brought in many fixes to the 1989 version and it has features that other software still omits (it is particularly good at deriving standard deviations from reports that do not directly provide them). The plan is for DSTAT 2 (see separate discussion) to incorporate these features. A 1997 version (1.20) brought in still more fixes:

  • Johnson, B. T. (1997). DSTAT 1.20: Software for the meta-analytic review of research literatures.Mahwah, NJ: Erlbaum

In all these versions, DSTAT 1.x has three effect-size derivation features that I no longer recommend because they are too conservative or too-error prone:

  1. Calculating effect sizes from significance levels (i.e., F10-S).
  2. Calculating effect sizes from proportions (i.e., F10-P)
  3. Many of the within-subject calculations are suspicious; procedures advanced by Becker (1988) have been shown to be superior. That is, using the pretest standard deviation is a superior solution to using the standard deviation of the paired comparisons. DSTAT 1.x uses the latter.

It has one analysis feature that I would no longer recommend using either, as DSTAT 2 offers more stable moderator testing and more options regarding assumptions:

  1. Continuous moderator testing (i.e., F7-C).

All of the analysis features in DSTAT 1.x operated under fixed-effects assumptions.

The program will also run under 64-bit systems (e.g., Windows XP, Vista, and System 7) if you use a 32-bit emulator such as dosbox. If you save dstat 1.1 to the directory c:\dstat you can work with it in dosbox using the following steps:

  1. Install dosbox if not already installed.
  2. Run it from the Windows Start menu.
  3. Mount the directory where you saved dstat.exe. That is, at the Z:\> prompt enter:
    1. mount c c:\dstat
    2. C:
    3. dstat
  4. Now that dstat is running, you can configure it and use it (these commands worked as of 18mar2011).

If you want a copy of DSTAT 1.x software and documentation, please send me an email.

Views: 1303

Replies to This Discussion

Hi all,

I installed dosbox and dstat 1.1 yesterday. I had to physically create a folder ("dstat") on my C drive ("Computer>System Volume (C:)") then, in dosbox, I had to type "mount c c:\dstat\", hit enter, type "C:", hit enter, and type "dstat" then everything was working fine for me. I had a bit of trouble at first, so I figured I'd share my success!! I believe the issues were creating a specific "dstat" folder, and including the backslash at the end of the mount command.

Jessa

Installed it today and it worked like a charm!  Thanks...

Hi fellow meta analyzers,

Here are some instructions I managed to put together for anyone interested in using DSTAT 1 in a Mac environment. Have fun!

Rob

Attachments:

Thanks Rob! -Blair

Hi all,

I came across an issue with DSTAT1.1 that I wanted to share.

Under the F test command [Different (but related) F value to g], I entered an F value with 6 decimal places (.000000). When I came to the end of the sequence, instead of being brought to the analysis page, I was reverted back to the main menu. 

After a few iterations, I found that the program will only work with 5 or less decimal places entered in a value field. I'm not sure if this holds for other areas besides the F-value entry, but something to keep in mind when computing ES.  

Best,

Lauren

Hi Lauren, when you say you entered an F-value with 6 decimal places, it looks like you entered 0.000000. F values can be of little use for reconstructing the SDpooled in ANOVA designs unless they differ from 0 somewhat. I'd be curious to see more of the problem that led you to enter an F of (apparently) exactly zero. Best, Blair

Hi Blair, 

The F value was actually 4.058981. I entered it as 6 zeros just to illustrate the place values. Sorry for the confusion!

Best, 

Lauren

Hmmm, I have not seen this problem before. But I just tried it in a couple of places and it worked fine. (See example attached.) Maybe when I see you tomorrow you could re-create the problem?

At least we know that meta-analysts do not routinely have 6 decimal places to enter! -Blair

Attachments:

Dear Blair, Please tell me how to get the a copy of DSTAT 1.x software and documentation.

Please email me.

Yay~~~~!!  Success!!!  Thank you a million Jessa!!!

Mark

Jessica M. LaCroix said:

Hi all,

I installed dosbox and dstat 1.1 yesterday. I had to physically create a folder ("dstat") on my C drive ("Computer>System Volume (C:)") then, in dosbox, I had to type "mount c c:\dstat\", hit enter, type "C:", hit enter, and type "dstat" then everything was working fine for me. I had a bit of trouble at first, so I figured I'd share my success!! I believe the issues were creating a specific "dstat" folder, and including the backslash at the end of the mount command.

Jessa

RSS

© 2017   Created by Blair T. Johnson.   Powered by

Badges  |  Report an Issue  |  Terms of Service