Databases
StackOverflow2013
Postsdbo
POHow to increase the number of permutations in the GLM cross validation function `cv.glm`
 

Note that there are some explanatory texts on larger screens.

plurals
  1. POHow to increase the number of permutations in the GLM cross validation function `cv.glm`
    primarykey
    data
    text
    <p>I'm wondering if anyone has any experience with specifying a number of permutations in the GLM cross validation function <code>cv.glm</code> (package: boot)? I am using the settings <code>K=2</code> for splitting my training and validation sets into approximately two equal groups. My <code>$delta</code> results can vary widely in some cases depending on the random seed. From the <code>?cv.glm</code> help, I cannot see any option for increasing the number of permutations. </p> <h3>Example</h3> <pre><code>require(boot) DF &lt;- structure(list(Y = c(0.158507483, 0.008510161, 0.002684648, 0.009587276, 0.001803681, 0.010173461, 0.002273384, 0.00345826, 0.051424454, 0.029937484, 0.194813452, 0.042138033, 0.022944148, 0.729585218, 0.887009621, 0.008899131, 0.001588576, 0.0216036, 0.001409499, 0.161051383, 0.026504919, 0.001495132, 0.059066545, 0.008317594, 0.490868633, 0.057027831), X1 = c(0.0974369543591941, -0.11971810600977, -0.168908964300336, -0.0011723143713434, -0.200018273737778, 0.0536459384966756, -0.188248143615029, -0.154736748196712, 0.0529959236206016, -0.152396350558232, 0.103766445240172, -0.0693365907826557, -0.114615555500542, 0.488829422819801, 0.561719898192691, -0.0469180067616361, 0.0631502939411764, -0.135689617930714, 0.0343957489602316, -0.0749974069726867, -0.107592097416425, 0.067741017650224, -0.167713403634508, 0.275062271178857, 0.276065626134302, -0.0926000525628916), X2 = c(-0.19192408577628, 0.116576354094024, 0.208731289320505, -0.138772290234524, 0.364065047213473, -0.1574052089755, 0.285540178523006, 0.29343767019163, -0.203222931158516, 0.0835579872715545, -0.157325117354138, -0.0242157560597033, -0.175123479037643, 0.174087353210292, 0.246559485637939, -0.43074835446357, -0.0181308378901971, 0.0525230701557242, -0.121813588478372, -0.0549274842561502, -0.115591654073407, -0.0190993986035446, 0.124566313208749, 0.138414677580375, -0.0981459346380045, -0.319191657096572 )), .Names = c("Y", "X1", "X2"), class = "data.frame", row.names = c(NA, -26L)) fmla1 &lt;- formula(Y ~ X1 + X2) glm1 &lt;- glm(fmla1, DF, family=gaussian(link="log")) summary(glm1) set.seed(111) cv1 &lt;- cv.glm(DF, glm1, K=2) cv2 &lt;- cv.glm(DF, glm1, K=2) cv3 &lt;- cv.glm(DF, glm1, K=2) cv4 &lt;- cv.glm(DF, glm1, K=2) cv1$delta; cv2$delta; cv3$delta; cv4$delta # RESULTS #[1] 0.007317702 0.005484949 #[1] 0.12918099 0.06621125 #[1] 1.029601e+31 -3.602880e+16 #[1] 0.02860412 0.01581949 </code></pre>
    singulars
    1. PO
      singulars
      1. PTAnswer
    1. This table or related slice is empty.
    1. PTQuestion
    1. USMarc in the box
    1. USMarc in the box
    plurals
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. PO
      singulars
      1. PTAnswer
    1. VO
      singulars
      1. POHow to increase the number of permutations in the GLM cross validation function `cv.glm`
      1. USMarc in the box
      1. VTBountyStart
    1. COBy "permutations", do you mean that you want to run the 2 fold cross-validation multiple times? If so, there is no option in `cv.glm` to do that, but it would be very easy to implement, even with just a loop. However, have you considered using more folds if you want to estimate variability? Have you considered a bootstrap (which seems particularly useful for what you want to do)?
      singulars
      1. POHow to increase the number of permutations in the GLM cross validation function `cv.glm`
      1. USnograpes
    2. COI didn't realize that you had only 26 points. If this is your real data, then you are going to get very poor convergence on models fit to 13 points.
      singulars
      1. POHow to increase the number of permutations in the GLM cross validation function `cv.glm`
      1. USnograpes
    3. CO@nograpes - Thanks for your efforts here. I think you've got the right approach below. Although this is not the right place to ask this, do you know of any rules of thumb for these `K` settings - i.e. How one chooses between the extremes of 2-fold cross validation vs. "leave-one-out" given the number of samples in the model?
      singulars
      1. POHow to increase the number of permutations in the GLM cross validation function `cv.glm`
      1. USMarc in the box
 

Querying!

 
Guidance
A row detail

Detail views are divided into sections. All the information in the data section comes from columns in the selected row. The other sections display data from other, related rows.

Related data can be related in a to-one or a to-many fashion. Captions of data related in a to-many fashion link to a list view showing a filtered view of the table.

Try moving around until you find a non-empty to-many entry and click on the label to get to one. You can move back to the root by clicking on the database name in the header.

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload