tetrad
tetrad copied to clipboard
cmu-phil
Handle missing value
When I try to process dataset with missing value, I get this:
Cutoffs should be in nondecreasing order: [0.0, NaN]
Cutoffs should be in nondecreasing order: [-6.378681, NaN]
algo mode:
--algorithm fges --parallelized --data-type mixed --delimiter comma --missing-marker NULL --score cg-bic-score --numCategories 16 --verbose
I guess maybe it is bug when processing dataset with missing value
Can you impute the missing values? In the Tetrad GUI testwise deletion is used for this case, though it is very slow. I supposed I could export that to tetrad-cmd...
In the Tetrad GUI testwise deletion is used for this case, though it is very slow. I supposed I could export that to tetrad-cmd...
I thought tetrad-cmd has this feature. I have done a impute version, I'm trying this to see if I can get better result
No, I looked at it but there are no descriptions.
On Wed, Dec 7, 2022 at 2:39 PM Joseph Ramsey @.***> wrote:
Did you ever try this? @lockmatrix https://github.com/lockmatrix
— Reply to this email directly, view it on GitHub https://github.com/cmu-phil/tetrad/issues/1405#issuecomment-1341494526, or unsubscribe https://github.com/notifications/unsubscribe-auth/AD4Y3OOFQWD7BFRQYEKB5YTWMDRVFANCNFSM57EPSNDQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Oh I meant that for the guy who posted the report... he said he was going to try something...
Joe
On Wed, Dec 7, 2022 at 3:19 PM cg09 @.***> wrote:
No, I looked at it but there are no descriptions.
On Wed, Dec 7, 2022 at 2:39 PM Joseph Ramsey @.***> wrote:
Did you ever try this? @lockmatrix https://github.com/lockmatrix
— Reply to this email directly, view it on GitHub <https://github.com/cmu-phil/tetrad/issues/1405#issuecomment-1341494526 , or unsubscribe < https://github.com/notifications/unsubscribe-auth/AD4Y3OOFQWD7BFRQYEKB5YTWMDRVFANCNFSM57EPSNDQ
. You are receiving this because you are subscribed to this thread.Message ID: @.***>
— Reply to this email directly, view it on GitHub https://github.com/cmu-phil/tetrad/issues/1405#issuecomment-1341540455, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACLFSR4C6HETMYDTUJPEWP3WMDWMTANCNFSM57EPSNDQ . You are receiving this because you commented.Message ID: @.***>
Did you ever try this? @lockmatrix
oh, I meant : I had already test a impute version. I was trying directly input the nullable data to see if I can get better result.
And I failed, got these message.
Ah. OK. I'll try to think about it some more then.
Thanks,
Joe
On Wed, Dec 7, 2022 at 8:10 PM Locke @.***> wrote:
Did you ever try this? @lockmatrix https://github.com/lockmatrix
oh, I meant : I had already test a impute version. I was trying directly input the nullable data to see if I can get better result.
And I failed, got these message.
— Reply to this email directly, view it on GitHub https://github.com/cmu-phil/tetrad/issues/1405#issuecomment-1341831751, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACLFSR2HVJV75EC2WYHXRUTWMEYRZANCNFSM57EPSNDQ . You are receiving this because you commented.Message ID: @.***>
Sorry to response late. Unfortunately, causal-cmd does not have the function to impute data, even though tetrad-gui has that function.
@lockmatrix We just did a new release of Tetrad 7.1.3-1; the corresponding causal-cmd 1.5.0 will follow soon. But I was going back through issues and realized what the issue is here. Sorry, I wasn't thinking before. In the interface, if you have missing values, the testwise deletion API will kick in and do its thing. But that's not available in causal-cmd.
Could it be made available in causal-cmd? That I'd have to think about; we would have to change the way it's being handled in the GUI and make --testwise-deletion a flag for all scores and algorithms that can handle it. That would be a pretty big job, and we're working with a skeleton crew at the moment (mostly me, with some help from kind people). But what I'll do here for now is close out this issue and open another issue to remind me to try to think about that further. If a way occurs to me to do it more easily than what I've got in mind currently, I may give it a shot. -JR