turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- Factor analysis variables

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

06-24-2017 08:45 AM

Hello

Since the variables like Safety, Usability, Puchase Experience,Contact experience and Look are discrete variables, I wonder how can we convert them into two factors using factor analysis as it is defined for continuous variables.

Accepted Solutions

Solution

07-13-2017
01:01 AM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-02-2017 07:18 AM

As I said earlier, you can't just say you want to use PROCEDURE X without saying why. This time, instead of saying you want to use PROCEDURE X, you say you want to reduce dimensions, but again the discussion of why you want to do this is completely missing.

Why do you want to reduce dimensions? What is the benefit in this case? What do you want to learn from this data? Is it really necessary to combine 5 variables into 2 dimensions?

Can you also explain how discrete variables become linear?

I'm really asking for a problem statement, not what procedure you want to use, or what step you plan to use in the analysis. Explain it like you are writing the abstract of a technical publication. Explain so someone who doesn't understand your problem can understand your problem.

As far as recommending books, since I don't understand the problem, I have no recommendation, other than, as discussed before, maybe you should consider using Correspondence Analysis, in which case you can certainly find books on this topic using your favorite search engine.

All Replies

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

06-24-2017 10:50 AM

While this does not literally answer your question, you can perform a principal component analysis of categorical data by using proc prinqual (sas/stat). Another alternative is correspondence analysis (proc corresp).

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

06-24-2017 02:29 PM

I think it is pointless to force the data into a continuous variable just so you can do a Factor Analysis. Although you CAN do this, it makes no sense. I believe that @WarrenKuhfeld has a good point, since these are discrete variables, then use a technique that works on discrete variables ... it is called Correspondence Analysis.

But in your diagram above, if you can somehow create Factor 1 and Factor 2, then what? What analysis do you really want to do with these 2 factors? I don't even see the direction you are trying to go in, much less what analysis you'd want to do. You can't start with a goal of using PROCEDURE X ... you have to first say what you want to learn from this information, and that is a mystery, you haven't explained what you want to learn from this information.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-02-2017 05:19 AM

Thanks for your valuable inputs.

I am try to reduce the dimension by combining the variables into two factors as mentioned in the image.

Also, after extracting factor1 and factor2, can we say that they are linear in nature?.

Can you please suggest me some books or case study link so that I can learn better.

Solution

07-13-2017
01:01 AM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-02-2017 07:18 AM

As I said earlier, you can't just say you want to use PROCEDURE X without saying why. This time, instead of saying you want to use PROCEDURE X, you say you want to reduce dimensions, but again the discussion of why you want to do this is completely missing.

Why do you want to reduce dimensions? What is the benefit in this case? What do you want to learn from this data? Is it really necessary to combine 5 variables into 2 dimensions?

Can you also explain how discrete variables become linear?

I'm really asking for a problem statement, not what procedure you want to use, or what step you plan to use in the analysis. Explain it like you are writing the abstract of a technical publication. Explain so someone who doesn't understand your problem can understand your problem.

As far as recommending books, since I don't understand the problem, I have no recommendation, other than, as discussed before, maybe you should consider using Correspondence Analysis, in which case you can certainly find books on this topic using your favorite search engine.