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
- /
- Health Care and Pharma
- /
- SAS in Health Care Related Fields
- /
- Estimate missing correlation coefficient in a corr...

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
- Permalink
- Email to a Friend
- Report Inappropriate Content

09-05-2008 08:59 AM

Good morning,

I have a scenario where i have been given a correlation matrix and in it i have many of the missing data. I would like to seek all of your help to please help me to figure out how could i do my best estimate in predicting those missing values in my correlation matrix.

I also would like to add that the correlation matrix contains non normalized data.

I have read SAS publication and i saw it offers an multiple imputation method for dealing with missing data. My question is :

1) Is IM dealing with only normalized data, but in my case, the matrix contains non normalized data.

2) Is IM only dealing with missing data, but in my case, I don't have a simple case of missing data, I have missing correlation coefficient in a correlation matrix. So could i use IM?

Thanks alot for all of your help,

Minh

I have a scenario where i have been given a correlation matrix and in it i have many of the missing data. I would like to seek all of your help to please help me to figure out how could i do my best estimate in predicting those missing values in my correlation matrix.

I also would like to add that the correlation matrix contains non normalized data.

I have read SAS publication and i saw it offers an multiple imputation method for dealing with missing data. My question is :

1) Is IM dealing with only normalized data, but in my case, the matrix contains non normalized data.

2) Is IM only dealing with missing data, but in my case, I don't have a simple case of missing data, I have missing correlation coefficient in a correlation matrix. So could i use IM?

Thanks alot for all of your help,

Minh

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

Posted in reply to deleted_user

10-21-2008 12:05 AM

Hi Minh,

Correlation matrix is a simetric matrix. Do you have the same missing values on both sides of the main diagonal?

Correlation matrix is a simetric matrix. Do you have the same missing values on both sides of the main diagonal?

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

Posted in reply to deleted_user

10-21-2008 12:07 AM

.

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

Posted in reply to deleted_user

10-21-2008 12:08 AM

.

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

Posted in reply to deleted_user

10-21-2008 12:12 AM

.

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

Posted in reply to deleted_user

10-21-2008 01:39 PM

The SAS PROC MI works only with missing data in the data table, it won't help you if you are starting with a correlation matrix and do not have the originating data. In fact, I don't think that there is a way to impute the missing correlations without the raw data.

If you have the raw data, then PROC MI can impute the missing data. It makes assumptions about the mechanism for the missingness (see the documentation for a discussion of that), but none about the underlying distribution in the data.

Doc Muhlbaier

Duke

If you have the raw data, then PROC MI can impute the missing data. It makes assumptions about the mechanism for the missingness (see the documentation for a discussion of that), but none about the underlying distribution in the data.

Doc Muhlbaier

Duke