[jira] [Commented] (SPARK-18501) SparkR spark.glm error on collinear data

["Apache Spark (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-18501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15676291#comment-15676291 ] 

Apache Spark commented on SPARK-18501:
--------------------------------------

User 'yanboliang' has created a pull request for this issue:
https://github.com/apache/spark/pull/15930

> SparkR spark.glm error on collinear data 
> -----------------------------------------
>
>                 Key: SPARK-18501
>                 URL: https://issues.apache.org/jira/browse/SPARK-18501
>             Project: Spark
>          Issue Type: Bug
>          Components: ML, SparkR
>            Reporter: Yanbo Liang
>            Assignee: Yanbo Liang
>
> Spark {{GeneralizedLinearRegression}} can handle collinear data since the underlying {{WeightedLeastSquares}} can be solved by local "l-bfgs"(rather than "normal"). But the SparkR wrapper {{spark.glm}} throw errors when fitting on collinear data:
> {code}
> > df <- read.df("data/mllib/sample_binary_classification_data.txt", source = "libsvm")
> > model <- spark.glm(df, label ~ features, family = binomial(link = "logit”))
> > summary(model)
> Error in `rownames<-`(`*tmp*`, value = c("(Intercept)", "features_0",  :
>   length of 'dimnames' [1] not equal to array extent
> {code}
> After depth study of this error, I found it was caused the standard error of coefficients, t value and p value are not available when the underlying {{WeightedLeastSquares}} was solved by local "l-bfgs". So the coefficients matrix was generated failed.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org