multivariate analysis
Locked
 
|
Hello. I am Thanasis. I want to do a statistical analysis for my work. I have around 53 variables in total. with spss i want to do a multivariate analysis. My purpose is to find and choose the most important variables for my statistical analysis. which method could you suggest me?manova method maybe? could you help me?
|
Administrator
|
Answering this question without knowing the nature of the research would be extremely irresponsible.
You need to throw a little more meat on the bones and sketch out some details. Number of cases, type of variables (measurement level, content...)ns? Do you have any sort of hypothesis or theoretical basis for making any sort of predictions? What sort of prior analysis background do you possess (stats and research design courses etc). Perhaps consider hiring a statistical consultant and/or reading a book or two about multivariate analysis. Or you could just throw everything in blindly and get the infamous GIGO model.
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
|
|
the truth is that my knowledge around the theory of multivariate analysis are not enough. I made a questionnaire with subject ''bullying at schools'' with 21 questions . I have been asked from my professor, to reduce the number of variables. I read from a book that multivariate analysis gives you the capability of a better analysis. Specifically i learn that multivariate analysis gives the possibility of condensing the information, containing many variables at least.there are many techniques multivariate analysis. The problem is that on spss i don't know how to do that. As i told you earlier i dont have knowledge around this theory and i understand that is difficult to you to help me. So if you could give me some advices(with steps for what is better to do) or send to me a link, to read the theory and application in spss of one technical Multivariate analysis, it could help. thank you for your time.
|
|
Email small number of cases of your SPSS input file to me to take a look. Also email what exactly your prof. asked to do on this case study (his own exact words is better) Max. From: thanosspan [via SPSSX Discussion] [mailto:[hidden email]] the truth is that my knowledge around the theory of multivariate analysis are not enough. I made a questionnaire with subject ''bullying at schools'' with 21 questions . I have been asked from my professor, to reduce the number of variables. I read from a book that multivariate analysis gives you the capability of a better analysis. Specifically i learn that multivariate analysis gives the possibility of condensing the information, containing many variables at least.there are many techniques multivariate analysis. The problem is that on spss i don't know how to do that. As i told you earlier i dont have knowledge around this theory and i understand that is difficult to you to help me. So if you could give me some advices(with steps for what is better to do) or send to me a link, to read the theory and application in spss of one technical Multivariate analysis, it could help. thank you for your time. If you reply to this email, your message will be added to the discussion below: http://spssx-discussion.1045642.n5.nabble.com/multivariate-analysis-tp5730087p5730089.html To start a new topic under SPSSX Discussion, email [hidden email] View this message in context: RE: multivariate analysis 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 thanosspan
To add to what MaxJasper said, walk us through this project beginning with the purpose of the study, the sample, especially how many people were surveyed and the questions this study is to provide data about. In your initial posting you said you had 53 variables. Here you say the survey had 21 questions. Explain that.
Gene Maguin -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of thanosspan Sent: Wednesday, July 08, 2015 7:39 PM To: [hidden email] Subject: Re: multivariate analysis the truth is that my knowledge around the theory of multivariate analysis are not enough. I made a questionnaire with subject ''bullying at schools'' with 21 questions . I have been asked from my professor, to reduce the number of variables. I read from a book that multivariate analysis gives you the capability of a better analysis. Specifically i learn that multivariate analysis gives the possibility of condensing the information, containing many variables at least.there are many techniques multivariate analysis. The problem is that on spss i don't know how to do that. As i told you earlier i dont have knowledge around this theory and i understand that is difficult to you to help me. So if you could give me some advices(with steps for what is better to do) or send to me a link, to read the theory and application in spss of one technical Multivariate analysis, it could help. thank you for your time. -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/multivariate-analysis-tp5730087p5730089.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 ===================== 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 MaxJasper
I sent to my professor to explain to me what exactly ASKS. and waiting for an answer .When he answer to me, I will send you the input to you to have a look.
ask something else: for ordinal variables ,Descriptive statistics I do are the percentiles and the median only; for nominal variables rates & mode? |
|
|
In reply to this post by thanosspan
If you don't have one or two specific hypotheses based on prior knowledge and
=====================
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
experiment, then you likely have an "exploratory study". But even for an exploratory study, you should want to reduce the number of important Scores so that the test you run will represent useful Hypotheses, without a vast proliferation of the number of tests. (See: multiple-test problem.) Literature review is useful: What has seemed useful in the past? The 30 items that are not part of the Bullying questionnaire should be reduced to a handful for the "primary analyses". The other variables will be used to characterize the whole sample, or, perhaps, eventual subgroups. Or they can be used to look at data in a more widely, exploratory mode. Frequencies and cross-tabs can be suggestive, or scatter-grams between continuous variables. Age and Sex probably stand alone, though the Age might contribute to another score for someone who is older and larger than his classmates. Given your set of data, WHAT do you hope or expect, at best, to be able to conclude? Factor analysis can be useful for "data reduction" of a scale. For example, I would suggest that 21 self-report items on Bullying should be reduced to (self-reported) totals for Victim, Aggression, Observed. However, if you do not have 5 or (much better) 10 times the number of subjects as you have items, a factor analysis is not necessarily reliable. On the other hand, eye-balling the questions very often gives obvious sets that you expect to go together, in which case you can start with these "pragmatic" factors based on apparent content; you can somewhat validate these by showing the all the items are positively correlated with each other. The thing to be alert for, assuming the scale is not well used in the past, is that particular items might be ambiguous, either because they were badly written, or because they were read differently by your particular sample. Hope this helps. -- Rich Ulrich > Date: Wed, 8 Jul 2015 16:38:53 -0700 > From: [hidden email] > Subject: Re: multivariate analysis > To: [hidden email] > > the truth is that my knowledge around the theory of multivariate analysis are > not enough. I made a questionnaire with subject ''bullying at schools'' with > 21 questions . I have been asked from my professor, to reduce the number of > variables. I read from a book that multivariate analysis gives you the > capability of a better analysis. Specifically i learn that multivariate > analysis gives the possibility of condensing the information, containing > many variables at least.there are many techniques multivariate analysis. The > problem is that on spss i don't know how to do that. As i told you earlier i > dont have knowledge around this theory and i understand that is difficult to > you to help me. So if you could give me some advices(with steps for what is > better to do) or send to me a link, to read the theory and application in > spss of one technical Multivariate analysis, it could help. thank you for > your time. > > |
|
|
In reply to this post by thanosspan
For descriptive statistics: A starting point is to pay attention to what other people in your area actually *report* -- You most likely should report approximately the same, when it comes to the ordinary variables. For age, the useful facts are usually the range (minimum, maximum) and the mean. However, if that "doesn't tell the whole story", ("maximum age was 75, but there were only 3 subjects over 50") then you owe your audience the fuller story. You can't begin to do a half-decent job of reporting on data if you have never, ever read (and noticed some details) in studies in your area. Your questions suggest that you need to do that first step. What you look at for your own information should be broader than what, in the end, will be worth reporting. Look at everything; report what is central, whatever is highly convention, and whatever is unusual in your own data-set. Look at the frequency counts for every variable -- For one thing, you should do that, just to make sure that nothing weird is going on... impossible values, or combinations that are unlikely. For instance, if there are a whole bunch of "0" or "88" for something like Age, it is probably Missing's that have not been accommodated. If two variables really *ought* to be similar, it should be heart-warming to prove it to yourself with a scatter-gram. If a nominal variable is worth using in an analysis, it is very conventional to report the counts for every category. The exception to that is when you do combine small-counts to get fewer categories -- the audience may or may not have an interest in those steps. Continuous variables need the mean and SD, and often, range; look for examples in GOOD JOURNALS or in much-cited articles in your literature. -- Rich Ulrich > Date: Thu, 9 Jul 2015 08:47:30 -0700 > From: [hidden email] > Subject: Re: multivariate analysis > To: [hidden email] > > I sent to my professor to explain to me what exactly ASKS. and waiting for an > answer .When he answer to me, I will send you the input to you to have a > look. > > ask something else: for ordinal variables ,Descriptive statistics I do are > the percentiles and the median only; for nominal variables rates & mode? > > |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |