[Daniel Bela]

I also opt for an asversion() option to saveold.

In addition to possible solutions discussed here, Jonathan may have a look at Sergiy Radyakin's -savespss-. I found a very convenient tool in it, and you get rid of any third-party software to create SPSS Statistics data files.
There seem to be some minor glitches, especially when very long value labels are involved. But this seems to be fixable in the future.

Regards
Bela

[Svend Juul]

Two user-generated programs may be helpful here:

Sergiy Radyakin wrote use13 which lets Stata 10-12 read files saved by Stata 13.
Lars Ängquist wrote use13save12 which can do a few more things. It also points to use10save9.

Jonathan Kelley isn't the first person to be confused and frustrated over StataCorp's re-re-recycling of the saveold command. I don't know if StataCorp has had thoughts whether to use a stick or a carrot strategy (stick: force old users to purchase new Stata version; carrot: being helpful both to new and old users). Hopefully the carrot will win; I also believe it is the wisest strategy in the long run.

[Nick Cox]

Several quite different issues are mixed up in this thread and already we have seen that some experienced users are confused (even misinformed!).

StataCorp as we all should know are in business selling Stata and that affects what they do. It's for them to consider what functionality might be added. But StataCorp want to encourage people to upgrade, which means sales. Ironically, the more that loyal users upgrade, the more difficulties those loyal users then have in collaborating with people using other software or much older versions of Stata. Naturally, the StataCorp view is that those other people are potential customers. But we all face limits. For example, some people don't believe that I can't read recent Excel files on a machine I often use but I have no interest in spending money to make that possible.

That aside -- although it's the key issue for many -- the requests of what StataCorp should do are strikingly asymmetric. People need to consider that being able to save to (much) older versions implies mangling your data increasingly as exporting increasingly older versions implies limits on variable types, number of observations, length of labels, representation of missing values, etc., etc. What are people asking for? That Stata 1 can read your data? That if somebody somewhere is still using Stata ? then that obliges StataCorp to make available saving datasets so that they can read them? Among other details, StataCorp must be coy of introducing functionality which will be misreported as Stata mangling your data when it saves to older versions. Urban myths can be bad for image.

I just checked the help for save (http://www.stata.com/help.cgi?save if you don't have Stata 13) and the purpose of saveold is explained as

Save data in memory to file in Stata 12 format

What's unclear about that?

On user-written programs, watch out. Sergiy's programs are a marked exception, as they do things inbuilt Stata commands don't offer. But some other user-written programs are just wrappers for saveold and don't extend its functionality in any fundamental sense. There is a history of even some programmers misunderstanding this, as a diligent search of the archives will uncover.

[Sergiy Radyakin]

Dear Prof. Kelley,

why would you prefer this complicated route when there exists a direct way?
Just install savespss.
With it's extensive list of supported features it is bound to be a tool of choice for almost any Stata to SPSS data transfer.

If you prefer to let SPSS do the import and you need the file in Stata 9 format, there are tools for that too.
Just install saveto9.

Both tools were advertized in Statalist (here and here) and in the SPSS forum.

Best, Sergiy Radyakin

[Richard Williams]

Sergiy, you need to check the forums more often. I knew about usespss but not the other programs.

-findit savespss- and -findit spss- do not find your program, as it is only available from your site and your site is not searched. Will you be adding it to SSC soon? Given that it is labeled as beta, are there issues to be concerned about?

[Nick Cox]

The database used by search, all (was: findit) is reconstituted daily. Not finding something may just mean that a site was down momentarily.

[Jazman Ihsanuddin]

Well, my short-term solution is saveold first in Stata 13, open it in Stata 12 and saveold again. *I'm serious*

[Richard Williams]

I have Stata 7 through 13 installed on my computer, so I think I am covered no matter what!

Just take seriously Nick's caution about mangling the data. If you are using data features that did not exist in earlier versions of Stata, then something has to give in the conversion process.

Sergiy's programs are great, especially if you are talking about tasks that only involve Stata and SPSS. But, you have to find them, learn how to use them, etc. We were nearly up to post 20 before his solutions were brought up. I still love Stat/Transfer because it is relatively cheap and can quickly do whatever I want and very little learning curve is involved.

Sergiy, if it doesn't exist already, it would be nice to have a single web page that briefly described all your file conversion utilities and then provided links to them. I put a list of suggested sites on my course pages, and I would add that page to the list if it existed. It would also be nice to get what you can on SSC -- if still beta then maybe just shout out some warnings -- because it is a lot easier to find something on SSC. Or, get your personal site included in the search results. Maybe you can even create a pseudo-entry for saveto9 that merely tells people how to get the program. Programs that are only mentioned in list messages are easy to overlook.

[Sergiy Radyakin]

Sergiy, you need to check the forums more often. I knew about usespss but not the other programs.

-findit savespss- and -findit spss- do not find your program, as it is only available from your site and your site is not searched. Will you be adding it to SSC soon? Given that it is labeled as beta, are there issues to be concerned about?

Dear Richard,

I do check the forum regularly, but it is not always that I can reply immediately. In some cases I need to get back to the office, in other I need to check that something works before getting it posted, etc, etc. And it takes time. Another problem is that most posters do not apply appropriate tags to their posts, and responders can't do that. Tags like "import, export, transfer, spss, file format" generally attract my attention .

As I wrote, both programs were advertized in the Statalist and a similar SPSS forum. I don't know how comfortable are the readers with every such advertisement. So it would help me, if somebody who is using programs that are already there starts threads with questions about them in Statalist. In most cases a solution or workaround can be found without any need to see or change the code, and can even be suggested by other users.

Google does find savespss. However Google gets confused, and tries to correct you, automatically presenting the results for "save spss" which is full of miscellaneous discussions, rarely going beyond literally repeating the manuals for the programs you/me/seeker don't have. Once you tell Google to search for "savespss" (spelled together) it will return the link to my page as the top result. Google has indexed the page immediately after the link was posted to the Statalist, and I do see a steady flow of users downloading -savespss- regularly (though mostly silently), and you can even see some users provided suggestions and bug reports still in 2013.

If you (or anyone) like the programs that I wrote, just tell your colleagues! spread the word! Facebook, Twitter, and other resources are now available for sharing the information like that. You will see on the page of use13, that when I encounter a university linking to my command, I may link back. But these are random, non-systemic finds, since I was not notified of them.

I do want to eventually post savespss to SSC and popularize it more. As a matter of fact -savespss- with another similar command was suggested for the Stata Conference in Boston this month, but was rejected in favor of other, no doubt, more useful commands and presentations. I do need to feel comfortable about the program's reliability before it gets out into the wild. However, since, e.g. SPSS data file format is not documented, the only way to feel comfortable is to run literally hundreds of different datasets through the program and check the output is valid. Doing this without an SPSS package literally means bugging other users to help out every time I make a change in my source. Similar problem is with -usespss-. An update to the 64-bit version was ready a few years ago, but I just haven't found yet a reasonably qualified (and equipped with SPSS) volunteer to provide independent testing. So I am not completely convinced it is worth replacing the 7-years old version which made it to a few books and had thousands of users with a version that only a few people who had interest have looked onto. Googling for usespss beta or usespss 64 will get you to the relevant page pretty fast though.

As for the potential problems in savespss, I try to document all features and all limitations on the homepage of savespss. Different users may put different weights on them. For example, I am not terribly concerned about dates. But some users might be. At the same time I find that reasonable coding of missing values is important, but Stat\Transfer behaves differently (not wrong, but differently). At this time strL support is probably the last feature I want to complete before releasing to SSC, but I wanted to do it in -use13- first.

Getting access to base version of Stata 10.0 would also be useful. Unfortunately StataCorp does not provide access to earlier versions of Stata, making it impossible for me to develop mata libraries compatible with some earlier versions of Stata. (The versions you get on the disk from StataCorp are not necessarily base versions. They are base versions with all the available updates applied, which means, for example, when you are buying Stata 13 now from StataCorp, you are getting Stata 13.1, and in general mata libraries compiled with it may not run in Stata 13.0, at least this is true for 10.0 and 10.1).

Best, Sergiy Radyakin

[Richard Williams]

Sergiy, thanks much. It is annoying how employers expect you to do things besides write Stata code. I have Stat/Transfer so I don't pay much attention to alternative programs, but yours are very useful for people who don't have convenient access to other choices. I have bookmarked this thread so I will better remember your programs in the future. If you have the time and inclination, I think it would be good to set up a page that lists all your programs and links to them, as that would make it easier to link to you.