turn on suggestions

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

Showing results for

- JMP User Community
- :
- Discussions
- :
- discriminant analysis

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
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Mar 16, 2017 10:33 AM
(1198 views)

Hi,

Can I get good performance in the Linear Discriminant Analysis (LDA), even if the normality assumption is violated ? Can I apply LDA in this condition? How can I justify that the LDA was performed in this way? I need some references to support this. (LDA was performed using two categories (one group of n = 41 and and other with n = 42).

Thank you

Adias

1 ACCEPTED SOLUTION

Accepted Solutions

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Mar 16, 2017 10:45 AM
(2309 views)

Solution

You MIGHT be able to get good performance in this situation, but you would be taking a risk. How much risk would depend on how far from normality your data are. Linear Discriminant Analysis is known to be sensitive to the normality assumption. Maybe you should consider logistic regression instead. That is an option that is not as sensitive to normality as LDA. Hosmer and Lemeshow is one reference that discusses the sensitivity to normality of LDA.

Dan Obermiller

2 REPLIES

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Mar 16, 2017 10:45 AM
(2310 views)

You MIGHT be able to get good performance in this situation, but you would be taking a risk. How much risk would depend on how far from normality your data are. Linear Discriminant Analysis is known to be sensitive to the normality assumption. Maybe you should consider logistic regression instead. That is an option that is not as sensitive to normality as LDA. Hosmer and Lemeshow is one reference that discusses the sensitivity to normality of LDA.

Dan Obermiller

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Mar 16, 2017 11:59 AM
(1183 views)

Thank you for your attention, answer and reference, that help me.

Adias