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
- :
- Discussions
- :
- predicted values in neural networks

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

Dec 22, 2011 5:44 AM
(5287 views)

Hi all,

can I use a "by" variable with n levels in neural network analysis?

I have already tried to insert a "by" variable with n levels in the analysis but I cannot save the predicted values for each level.

Thank you in advance for your reply

BR

Maria

1 ACCEPTED SOLUTION

Accepted Solutions

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

Hi Maria,

Thanks for using JMP! I'm sorry that you have been inconvenienced by this issue. It is a known issue that may be fixed in a future version of JMP.

In the meantime, here is a possible workaround that might help. I've used the Big Class sample data table to illustrate the concept.

<!

dt = open("$SAMPLE_DATA/Big Class.jmp");

nn = Neural(

Y( :weight ),

X( :height, :age ),

Validation Method( Holdback, 0.3333 ),

Fit( NTanH( 3 ), Save Formulas ),

By( :sex )

);

wait(0.5);

dtF = datatable( "sex=F" );

dtM = datatable( "sex=M" );

dtAll = dtF << Concatenate( dtM );

-->

-Michael

Message was edited by: Michael Crotty

Michael Crotty

Sr Statistical Writer

JMP Development

Sr Statistical Writer

JMP Development

3 REPLIES

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

Hi Maria,

Thanks for using JMP! I'm sorry that you have been inconvenienced by this issue. It is a known issue that may be fixed in a future version of JMP.

In the meantime, here is a possible workaround that might help. I've used the Big Class sample data table to illustrate the concept.

<!

dt = open("$SAMPLE_DATA/Big Class.jmp");

nn = Neural(

Y( :weight ),

X( :height, :age ),

Validation Method( Holdback, 0.3333 ),

Fit( NTanH( 3 ), Save Formulas ),

By( :sex )

);

wait(0.5);

dtF = datatable( "sex=F" );

dtM = datatable( "sex=M" );

dtAll = dtF << Concatenate( dtM );

-->

-Michael

Message was edited by: Michael Crotty

Michael Crotty

Sr Statistical Writer

JMP Development

Sr Statistical Writer

JMP Development

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

I like your method, but I wish the Predicted Value and Hidden Node columns would retain their formulas after the by groups are concatenated into one data table.

How would you retain column formulas in the concatenated data table?

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

Retaining column formulas in the concatenated data table would be tricky because they are not the same formulas in the multiple tables. A single column can only have one formula.

To get around this, you could create a formula with an If() statement where the "then" clauses are the by group formulas... Something like this:

If( :sex == "F", *pred formula "F"*, :sex == "M", *pred formula "M"* )

To do that, you'd probably do the concatenation of the tables and then create the formulas and add them to the Predicted Value and Hidden Node columns after the fact.

Michael Crotty

Sr Statistical Writer

JMP Development

Sr Statistical Writer

JMP Development