Error computing expected influence from case-dropping bootstrap about bootnet

[dpmoriarity]

I received the same error: " in data.frame(name = name, type = "expectedInfluence", node1 = x[["labels"]], :
arguments imply differing number of rows: 1, 18, 0") when running a case-dropping bootstrap. The nonparametric bootstrap ran fine. I was also using a custom function (code for function, network, and bootstraps below)

lasso_function <- function(data){EBICglasso(cor(data,use="complete.obs", method="spearman"), n = nrow(data))}

T1Data_S_network<- estimateNetwork(T1Data_S,default="none",fun=lasso_function)

b1 <- bootnet(T1Data_S_network, nBoots = 1000,statistics=c("edge","strength","expectedInfluence"), labels = labels_S, type = "nonparametric")
b2 <- bootnet(T1Data_S_network, nBoots = 1000, statistics=c("edge","strength","expectedInfluence"), labels = labels_S, type = "case")

Any help would be appreciated, thank you!

-Dan

[brklynn]

I can run nonparametric as well. It seems unique to the combination of case + expected influence. Hopefully this can get elucidated for us.

[mayakup]

Hi all,

I noticed I tend to receive this error message when I bootstrap a network that often returns empty. Did you come across a similar pattern?

[dpmoriarity]

I didn't personally get any warnings or errors to that effect when using estimatenetwork or during the bootstrapping procedure.

[dpmoriarity]

Actually, sorry about that, wrong model. Yes, the bootstrapping procedure is giving me warnings about empty networks being selected as the best fitting models (this came up eight times). This warning only comes up in the case dropping bootstrap.

[SachaEpskamp]

Hi all,

This should now be resolved in the devel version on Gtihub. Could anyone check? Then I'll submit it to CRAN.

library("devtools")
install_github("sachaepskamp/bootnet")

Best, Sacha

[dpmoriarity]

Thanks, Sacha. I'm still getting the same error. Not sure if this is useful information but I got this message while updating.

Error: Failed to install 'bootnet' from GitHub:
(converted from warning) package ‘foreach’ is in use and will not be installed.

[brklynn]

Sacha,

It worked for me!

Notably, when trying to download the Github version of bootnet, I also received many errors similar to what Dan is mentioning. For example: Error: Failed to install 'bootnet' from GitHub:
(converted from warning) cannot remove prior installation of package ‘backports’

How I fixed this was by installing each package it was trying to remove
e.g. install.packages("backports")

Once I did this for digest, Rccp, and backports, the bootnet download went through, and I could perform the case-dropping bootstrap of expected influence successfully.

Thank you very much! Please let me know if I can help further.

Brooklynn

[dpmoriarity]

That worked for me too! Cleared the workspace and installed the problematic package by itself. However, I'm getting a similar error (potentially due to a different modeling issue) for a later bootstrapping procedure: This time it having issues with the strength metric. Also, it comes up with both the nonparametric and case-dropping bootstraps. It differs from the last network in that it is dense, so I'm not sure if this is driving the issue or not.

Network with lowest lambda selected as best network: assumption of sparsity might be violated.
Computing statistics...
Error in data.frame(name = name, type = "strength", node1 = x[["labels"]], :
arguments imply differing number of rows: 1, 18, 13

[SachaEpskamp]

Thanks! I will close this issue then. Perhaps you can open a new issue about the new message? Do you get it when bootstrapping or when estimating the original network?