In a similar vein to the question Kirill just posted, I wonder what has happened to SPSS AnswerTree which I used extensively for some years in Market Research. It had so many more options than the TREE command.
Is AnswerTree no longer available? If so, when was it discontinued and why? TIA, Ruben |
Ruben,
Can you list some features available in Answer Tree and not in SPSS Statistics' Tree? It is interesting. Some people may have Answer Tree installed and never used it because they just thought it is identical to TREE command. |
I don't have AnswerTree anymore but I might gain access to it via a network license my ex colleagues may be using. Anyway, I think CHAID and CRT are not sound approaches to data analysis. I've been planning to write an article on this but I haven't found the time yet.
Anyway, in AnswerTree you could build the tree automatically just as with TREE. What would often happen, is that small subgroups would split off high in the tree or splits were made that didn't make sense. You could not present most automatic trees to a client. Now in AnswerTree you could correct this by right-clicking any node in the tree and selecting the second best (instead of first best) split for each node. Or the third (... fifth) best. Like so you could "manually" grow a tree without tiny subgroups or undesirable splits. In most cases you really needed these options. Of course, such a tree would be rather subjective but at least it would be sort of useful for your clients. HTH, Ruben |
In editing TREE output, you can hide splitting of any node you want.
I can't say anything about your last paragraph. But even if it is
not possible in TREE, it is an output deficiency, not an algorithm
deficiency.
28.05.2014 12:32, Ruben Geert van den
Berg пишет:
I don't have AnswerTree anymore but I might gain access to it via a network license my ex colleagues may be using. Anyway, I think CHAID and CRT are not sound approaches to data analysis. I've been planning to write an article on this but I haven't found the time yet. Anyway, in AnswerTree you could build the tree automatically just as with TREE. What would often happen, is that small subgroups would split off high in the tree or splits were made that didn't make sense. You could not present most automatic trees to a client. Now in AnswerTree you could correct this by right-clicking any node in the tree and selecting the second best (instead of first best) split for each node. Or the third (... fifth) best. Like so you could "manually" grow a tree without tiny subgroups or undesirable splits. In most cases you really needed these options. Of course, such a tree would be rather subjective but at least it would be sort of useful for your clients. HTH, Ruben -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/What-happened-to-SPSS-AnswerTree-tp5726236p5726238.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== 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 |
True, but put the "output deficiency" isn't my biggest concern with CHAID and CRT. However, I don't currently have the time to discuss other concerns thoroughly so I kinda have to leave it at that for now.
|
Ruben,
If you have an inclination share your concers about the two methods here. People might be willing to appraise it. |
In reply to this post by Ruben Geert van den Berg
AnswerTree was discontinued many years
ago, because the market for it given the tree capabilities in Statistics
and Modeler was very small. The Statistics TREE command has all of
the AT capabilities except for the ability to build or edit the tree interactively
except as below. That functionality would have been very difficult
to provide in the SPSS architecture, and it was found that not many AT
users used it. Controlling the splitting parameters seems to work
reasonably well even though the idea of interactive tree building seems
very appealing.
You can, in the Tree Editor, collapse/hide children of a node, but you can't steer the building process or merge a subset of the children of a node In Modeler, interactive tree building is still available for certain tree algorithms. Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] phone: 720-342-5621 From: Ruben Geert van den Berg <[hidden email]> To: [hidden email], Date: 05/28/2014 02:33 AM Subject: Re: [SPSSX-L] What happened to SPSS AnswerTree? Sent by: "SPSSX(r) Discussion" <[hidden email]> I don't have AnswerTree anymore but I might gain access to it via a network license my ex colleagues may be using. Anyway, I think CHAID and CRT are not sound approaches to data analysis. I've been planning to write an article on this but I haven't found the time yet. Anyway, in AnswerTree you could build the tree automatically just as with TREE. What would often happen, is that small subgroups would split off high in the tree or splits were made that didn't make sense. You could not present most automatic trees to a client. Now in AnswerTree you could correct this by right-clicking any node in the tree and selecting the second best (instead of first best) split for each node. Or the third (... fifth) best. Like so you could "manually" grow a tree without tiny subgroups or undesirable splits. In most cases you really needed these options. Of course, such a tree would be rather subjective but at least it would be sort of useful for your clients. HTH, Ruben -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/What-happened-to-SPSS-AnswerTree-tp5726236p5726238.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== 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 Kirill Orlov
@Jon: thank you, that was just what I was looking for!
@Kirill: some points you should only make after some careful research. Until then, no concerns here! |
Free forum by Nabble | Edit this page |