Using the Statistical Program R for Running Statistics

  • ashende (3/10/2014)


    Thank you for the article. It is nice to know that this tool is available and it may be useful in some cases. However, I think Excel would be a better tool to use in most cases.

    Just remember: The London Whale did his damage running an Excel spreadsheet (which he clearly didn't understand). The standard rejoinder is, "spreadsheets don't kill people, people kill people", but the fact is Excel simply doesn't have the horsepower to do things right. As a result, "analysts" adjust their work to the limitations of Excel. Perhaps we should call all analysts Jonah?

    (aside: here's a recap of Morgan's postmortem, from here: http://www.dangreller.com/the-london-whale-operational-maturity-and-microsoft-excel/ )

    The report points out the following issues with the use of these Excel spreadsheets:

    A critical model used to determine risk was based on a series of spreadsheets which were manually maintained by copying and pasting data between the sheets

    A spreadsheet error failed to reflect a $400 million loss

    A key spreadsheet used to determine risk used an incorrect statistical model for calculations. The spreadsheet had a cell that determined which model it would use and defaulted to the wrong model

    Another spreadsheet error made an incorrect calculation by dividing by the sum of two numbers, instead of dividing by their averages

    Inadequate tracking of changes to spreadsheets prevented personnel from determining when errors were first introduced

    Or, don't let your juveniles near the sharp cutlery.

  • I'm a big fan of eusprig - the horror stories are scary stuff but it also contains some useful best practice documents

    http://www.eusprig.org/[/url]

  • Or you could just use the built-in SQL Server data mining and DMX extensions without needing to use xp_cmdshell and learn R at all!

  • This is a great article that really allows you to integrate the R statistics in a production environment based on SQL SERVER.

    Thank You !

    Alessio M.

  • I am also no fan of the cmdshell method permissions requirement for running an R script. However, in SQL Server 2016, R Services enabled, with external scripts enabled, we can use sp_execute_external_script to run an R script from within sql mangler studio, and have results sent back.. all nicely integrated, as it should be. 😎

  • More info on R, thanks.

  • Great work Tomaz,

    Thanks for sharing this wonderful integration and opening up lots of possibilites with these two great products (R and SQL). Excellent work.

Viewing 7 posts - 16 through 21 (of 21 total)

You must be logged in to reply to this topic. Login to reply