Keilhacker, Michael
2017-Jun-05 10:14 UTC
[R] R] Error message "cs_lu(A) failed: near-singular A (or out of memory)"
Dear Arne, thank you very much upfront for your time that you invest in answering my question. I'm using your systemfit R package to analyze if companies risk appetite has a relationship to their performance and how it evolves over time. We collected data for ca. 80 companies, from 6 segments over 15 years that gives us a final data set with 1,312 firm-year observations. We have set up 3 equations (based on existing SEM from literature) and experience the following: - The systemfit package performs well when using no dummy variables (e.g. for year, segment, firm) - When adding dummy variables (year / segment) the systemfit package still provides results; however, if we remove the segment dummy variable and add a firm dummy variable we experience the error message: "cs_lu(A) failed: near-singular A (or out of memory)" - We also tested the systemfit package with firm and year dummy variable using a sub-sample of our data (ca. 10 companies) and the package worked without any error messages - Afterwards we tested the systemfit package using a server (64 gb ram, 16 core,...) by applying the systemfit package again on the final data set. However, we experienced again the same error message. We use R 3.3.2 and the systemfit package 1.1-20. I would be very grateful if you could support us anyhow and give us a hint how we could solve the situation. Thanks again for your help. Warm Regards from Munich, Michael ___________________________________________________ Michael Keilhacker, MBA, M.A. Logistics and Supply Chain Management Technische Universit?t M?nchen ? TUM School of Management Arcisstra?e 21 ? 80333 M?nchen ? +49-163-5454918<tel:%2B49-89-289-28203> ? michael.keilhacker at tum.de<mailto:christian.mandl at tum.de> www.log.wi.tum.de<http://www.log.wi.tum.de> [[alternative HTML version deleted]]
Jeff Newmiller
2017-Jun-05 16:40 UTC
[R] R] Error message "cs_lu(A) failed: near-singular A (or out of memory)"
This appears to be a case of insufficient understanding of statistics, which is not on topic for r-help. Look at how well the per-firm models would work... my guess is that some of the firms have insufficient data for useful analysis. If that is not it, consider providing a more concrete example on stats.stackexchange.com. -- Sent from my phone. Please excuse my brevity. On June 5, 2017 3:14:59 AM PDT, "Keilhacker, Michael" <michael.keilhacker at tum.de> wrote:>Dear Arne, > > >thank you very much upfront for your time that you invest in answering >my question. > > >I'm using your systemfit R package to analyze if companies risk >appetite has a relationship to their performance and how it evolves >over time. We collected data for ca. 80 companies, from 6 segments over >15 years that gives us a final data set with 1,312 firm-year >observations. We have set up 3 equations (based on existing SEM from >literature) and experience the following: > >- The systemfit package performs well when using no dummy variables >(e.g. for year, segment, firm) > >- When adding dummy variables (year / segment) the systemfit package >still provides results; however, if we remove the segment dummy >variable and add a firm dummy variable we experience the error message: >"cs_lu(A) failed: near-singular A (or out of memory)" > >- We also tested the systemfit package with firm and year dummy >variable using a sub-sample of our data (ca. 10 companies) and the >package worked without any error messages > >- Afterwards we tested the systemfit package using a server (64 gb ram, >16 core,...) by applying the systemfit package again on the final data >set. However, we experienced again the same error message. > > >We use R 3.3.2 and the systemfit package 1.1-20. > > >I would be very grateful if you could support us anyhow and give us a >hint how we could solve the situation. > > >Thanks again for your help. > > >Warm Regards from Munich, > >Michael >___________________________________________________ > >Michael Keilhacker, MBA, M.A. >Logistics and Supply Chain Management >Technische Universit?t M?nchen ? TUM School of Management >Arcisstra?e 21 ? 80333 M?nchen ? >+49-163-5454918<tel:%2B49-89-289-28203> ? >michael.keilhacker at tum.de<mailto:christian.mandl at tum.de> >www.log.wi.tum.de<http://www.log.wi.tum.de> > > [[alternative HTML version deleted]]