Back to PLS Helpbatch script in gui
Posted on 07/02/10 11:38:12
Number of posts: 41
Hi all ~
I have two questions...
1. Can you load the batch script for a PLS analysis into the gui? You can save out a batch script - but I am looking for the reverse. This is handy if one wants to check visually that all parameters look ok ...or switch to the gui.
2. Is there any rationale to justify removing an LV from the dataset and rerunning the PLS? (akin to ICA). I ask b/c I tend to find a common LV result, followed by an interesting (but non-significant) LV. I wonder if that second LV would be reliable enough if tested within the remaining "noise" after removing the first LV. I know my data are noisy (S/R was poor) - and so I wonder if my common effect is overshadowing any other subtleties...
Best ~
Agatha
I have two questions...
1. Can you load the batch script for a PLS analysis into the gui? You can save out a batch script - but I am looking for the reverse. This is handy if one wants to check visually that all parameters look ok ...or switch to the gui.
2. Is there any rationale to justify removing an LV from the dataset and rerunning the PLS? (akin to ICA). I ask b/c I tend to find a common LV result, followed by an interesting (but non-significant) LV. I wonder if that second LV would be reliable enough if tested within the remaining "noise" after removing the first LV. I know my data are noisy (S/R was poor) - and so I wonder if my common effect is overshadowing any other subtleties...
Best ~
Agatha
Replies:
Untitled Post
I'm Online
Posted on 07/05/10 11:28:03
Number of posts: 291
For your first question: I will look into it. It's a good suggestion anyway, thanks.
For your second question: we only obtain LV components after the PLS analysis; so there is no LV can be removed before we run the analysis. Usually, you can find the pattern of the LV reliability from the permutation test.
For your second question: we only obtain LV components after the PLS analysis; so there is no LV can be removed before we run the analysis. Usually, you can find the pattern of the LV reliability from the permutation test.
Untitled Post
I'm Online
Posted on 07/05/10 11:38:05
Number of posts: 229
2. Is there any rationale to justify removing an LV from the dataset
and rerunning the PLS? (akin to ICA). I ask b/c I tend to find a common
LV result, followed by an interesting (but non-significant) LV. I
wonder if that second LV would be reliable enough if tested within the
remaining "noise" after removing the first LV. I know my data are noisy
(S/R was poor) - and so I wonder if my common effect is overshadowing
any other subtleties...
Agatha - we often residualize the dominant LVs from the data when we want to plot the more subtle (smaller) effects in the data. Whether or not your "nonsignificant" LVs would become "significant" will depend in large part to the strength of LV1 relative to the remaining LVs. If LV2 is small (in the noise), you are unlikely to see any improvement by residualizing and rerunning the analysis - but it is certainly worth a try..
n
Agatha - we often residualize the dominant LVs from the data when we want to plot the more subtle (smaller) effects in the data. Whether or not your "nonsignificant" LVs would become "significant" will depend in large part to the strength of LV1 relative to the remaining LVs. If LV2 is small (in the noise), you are unlikely to see any improvement by residualizing and rerunning the analysis - but it is certainly worth a try..
n
Untitled Post
Posted on 07/05/10 14:10:11
Number of posts: 41
thank you both!
i'll check out the residualizing - i was initially surprised that you didn't laugh at my question - i thought it was actually a silly question given that LVs from a SVD are orthogonal already. but if the permutation distribution is dependent on the relative magnitude of the singular values then residualizing does have some promise. i would have actually thought that it makes most sense if LV2++ are a lot smaller than LV1, because the distribution of singular values would be the most affected by residualizing... i'll check it out - i am curious to see what happens. thanks nancy!
agatha
i'll check out the residualizing - i was initially surprised that you didn't laugh at my question - i thought it was actually a silly question given that LVs from a SVD are orthogonal already. but if the permutation distribution is dependent on the relative magnitude of the singular values then residualizing does have some promise. i would have actually thought that it makes most sense if LV2++ are a lot smaller than LV1, because the distribution of singular values would be the most affected by residualizing... i'll check it out - i am curious to see what happens. thanks nancy!
agatha
Login to reply to this topic.
