Loading ...
Sorry, an error occurred while loading the content.

Re: [XP] XP and Scrum

Expand Messages
  • Manuel Klimek
    Chris, On Wed, May 14, 2008 at 9:07 AM, Chris Wheeler ... 1. if you have a measurable dependent variable, which we haven t. 2. if you have enough consistent
    Message 1 of 138 , May 14, 2008
    • 0 Attachment
      Chris,

      On Wed, May 14, 2008 at 9:07 AM, Chris Wheeler
      <christopher.wheeler@...> wrote:
      > On Wed, May 14, 2008 at 11:48 AM, Manuel Klimek <klimek@...> wrote:
      >
      >>
      >> From what I know about maths, combining metrics about which we have no
      >> idea how strongly they are correlated with the dependent variable /and
      >> each other/ means that we can't say anything about it. The 'errors'
      >> might cancel out, they might multiply, and I think the basic problem
      >> is to find the correlation of the independent variables. If we knew
      >> these, we could create a metric that explains the dependent variable
      >> in a better way.
      >>
      >> I don't think regression analysis applies here because of the
      >> interdependence of the errors of the independent variables.
      >
      > Ok. Like I said, read up on it. There are ways to discern multicollinearity
      > amongst independent variables and ways to deal with it. Regression analysis
      > is helpful in determining how much variation in your dependent variable is
      > accounted for by your independent variables.

      1. if you have a measurable dependent variable, which we haven't.
      2. if you have enough consistent data that you can do sound
      statistical analysis, which is moot if you don't have (1), but which
      is hard to come by, even if you have (1). With consistent I mean that
      the environment does not change in a way that the metrics change
      without the target metric changing. Which would lead to all that
      repeatability that CMMI seems to be about, which seems to lead to
      making the same error over and over again, just to be able to prove
      that you made it.
      3. I said interdependence of the errors, which I think is something
      else than covariance of the independent variables, whilst related to
      it and probably computable if you could find out the covariances. But
      I confess that I am not on firm ground here.

      > Or, don't read up on it. Whatever suits you.

      I already read up on it since you hinted me to do so, and I knew about
      it before, having had some graduate math during CS and finance
      studies. Just thought I'd mention it :-)

      Cheers,
      /Manuel

      --
      http://klimek.box4.net
    • Manuel Klimek
      Chris, On Wed, May 14, 2008 at 9:07 AM, Chris Wheeler ... 1. if you have a measurable dependent variable, which we haven t. 2. if you have enough consistent
      Message 138 of 138 , May 14, 2008
      • 0 Attachment
        Chris,

        On Wed, May 14, 2008 at 9:07 AM, Chris Wheeler
        <christopher.wheeler@...> wrote:
        > On Wed, May 14, 2008 at 11:48 AM, Manuel Klimek <klimek@...> wrote:
        >
        >>
        >> From what I know about maths, combining metrics about which we have no
        >> idea how strongly they are correlated with the dependent variable /and
        >> each other/ means that we can't say anything about it. The 'errors'
        >> might cancel out, they might multiply, and I think the basic problem
        >> is to find the correlation of the independent variables. If we knew
        >> these, we could create a metric that explains the dependent variable
        >> in a better way.
        >>
        >> I don't think regression analysis applies here because of the
        >> interdependence of the errors of the independent variables.
        >
        > Ok. Like I said, read up on it. There are ways to discern multicollinearity
        > amongst independent variables and ways to deal with it. Regression analysis
        > is helpful in determining how much variation in your dependent variable is
        > accounted for by your independent variables.

        1. if you have a measurable dependent variable, which we haven't.
        2. if you have enough consistent data that you can do sound
        statistical analysis, which is moot if you don't have (1), but which
        is hard to come by, even if you have (1). With consistent I mean that
        the environment does not change in a way that the metrics change
        without the target metric changing. Which would lead to all that
        repeatability that CMMI seems to be about, which seems to lead to
        making the same error over and over again, just to be able to prove
        that you made it.
        3. I said interdependence of the errors, which I think is something
        else than covariance of the independent variables, whilst related to
        it and probably computable if you could find out the covariances. But
        I confess that I am not on firm ground here.

        > Or, don't read up on it. Whatever suits you.

        I already read up on it since you hinted me to do so, and I knew about
        it before, having had some graduate math during CS and finance
        studies. Just thought I'd mention it :-)

        Cheers,
        /Manuel

        --
        http://klimek.box4.net
      Your message has been successfully submitted and would be delivered to recipients shortly.