Lack of Robustness to Missing Values in Dependent Variable

[DarioS]

While looping through a bunch of outcomes, miloR faltered on a column that has some missing values.

> testNhoods(MILOdata, design = ~ Recurrence, design.df = design, fdr.weighting = "graph-overlap")
  Error in dimnames(x) <- dn: length of 'dimnames' [1] not equal to array extent
> head(design)
             Sample Recurrence
HN118_P     HN118_P       <NA>
OSCC_12-P OSCC_12-P        Yes
OSCC_16-P OSCC_16-P        Yes
OSCC_16-M OSCC_16-M        Yes
OSCC_20-P OSCC_20-P       <NA>
OSCC_22-P OSCC_22-P        Yes

I expect it could automatically handle it by avoiding any samples with a missing outcome value.

[MikeDMorgan]

Hi @DarioS - we leave it to the user to define how they wish to handle missing values.

[DarioS]

Ah. Do Milo objects have any special subsetting functions to remove certain samples?

[MikeDMorgan]

If you remove the samples you want to from the data frame you pass to design.df, then the subsetting on the nhoods count will happen automatically within testNhoods. As long as the rownames of your design data frame are a proper subset of the column names of nhood counts, it should work.