1 \documentclass{article
}
3 \title{CLS: an approach for a new statistical system
}
11 \section{Introduction
}
14 Statisticians who use a computer for data analysis invariably take one
15 of two approaches (considered in the extremes here for illustration):
17 \item the
\emph{FORTRAN
} approach of coding numerical and algorithmic
18 information into the computer program code used for the data
20 \item the
\emph{GUI
} approach, via Microsoft Excel, SPSS, Minitab, and
21 similar approaches, where tasks are facilitated, sometimes with
22 accompanying workflow support.
24 Both approaches have co-existed since the early
80s, with the FORTRAN
25 approach dating back to the dawn of the computing era.
27 \section{Components of a procedure
}
28 \label{sec:components
}
30 define a statistical procedure as a decision-making approach which
31 entails the intertwining of formal and informal structure.
35 \item \label{statproc-decision
} Decision to make
36 \item \label{statproc-assessment
} Assessment approach to use
37 (some are inherently different, others just look different)
38 \item \label{statproc-normalization
} Normalization of the problem for
39 assessment/comparison with other reference behaviours
40 \item \label{conclusion
} Type of conclusion desired, and instance of
41 that conclusion (when data is present)
44 This forms an
\textit{abstract class
} of a procedure, which can be
45 represented by a real class, which can then be instantiated through
46 the application of data.
49 \label{sec:components:decision
}
51 By example, consider the t-test as an instance of a procedure,
52 representing the general class of testing hypotheses surrounding
2
53 means. Related would be formal likelihood tests with distributions,
54 the superspace/classes from regression and ANOVA.
57 \item are the
2 means the same?
58 \item what is the difference?
59 \item what is the strength of the difference?
62 \subsection{Core Assessment
}
63 \label{sec:components:assessment
}
65 This is the construction of the model and parameters that would be
66 used to form the term used to make the assessment. Here, we could
69 \label{eq:assess:ex:
1}
70 \hat{E
}[Y|G=
1] -
\hat{E
}[Y|G=
0]
72 as the fundamental quantity to compare. This can arise from many
73 sources such as regression models
75 \label{eq:assess:ex:
2}
76 Y =
\mu +
\beta G +
\epsilon \\
81 \label{eq:assess:ex:
2}
82 E
[Y|G
] =
\mu +
\beta G
85 \subsection{Normalized Behavior
}
86 \label{sec:components:normbeh
}
87 Let $X=(Y,G)$ from above, the whole data.
92 \frac{ \hat\mu_1 -
\hat\mu_0}%
93 {\hat{SE
}(
\hat\mu_1 -
\hat\mu_0)
}
95 or regression-model-based:
101 or likelihood-model-based: (FIXME!)
104 -
2 \log \frac{ L(
\hat\beta|X)
}%
107 or score-model-based:
110 \cal{I
}^
{-
1}(
\beta=
0,X) S(
\beta=
0,X)
113 \subsection{Conclusion Desired
}
114 \label{sec:component:conclusion
}
116 Value or Range on the Target Scale (existing parameter describing
117 data-oriented substantive model)
119 Translation of Value/Range on the Decision Scale (what to do, what to
120 decide about the problem, i.e. in a testing framework).
122 \section{Class Implementation
}