Current loctn:
about >
methods >
data analysis

## Data Analysis

Initial stages of data analysis generally make heavy use of graphical data
displays. Histograms are generally not used for display of distribution of
values on an ordinal or nominal scale due to well-known problems with
obscuring of data features when inappropriate category limits are chosen.
Techniques employing empirical data distribution graphs are employed
instead.

Use of robust and non-parametric techniques is emphasized. If a
parametric analysis employing standard techniques such as least squares
regression is employed, an alternative analysis employing a robust technique
will generally be employed to verify the parametric results.

Assumptions underlying applied statistical tests are always tested for
violations.

Statistical analyses is generally performed in either R or SPSS statistical
software. Work can be completed using SAS if the client provides an
appropriate SAS usage license. Microsoft Access is frequently employed
for implementation of data transformations that are most efficiently
performed using structured query language (SQL). Perl, Lisp, Fortran,
Pascal, VisualBasic, MySQL, PostGreSQL may also be used if appropriate.

Documents providing short summaries of all major programs developed to
implement an analysis are maintained.

All programs are maintained in formal revision control systems.

Documents summarizing all analysis interpretations are maintained
throughout analysis. More concise reports are generally provided at the
end of the analysis or at important milestones. These reports are
targeted to assist the client in understanding the key results of the
analysis.

All documents and program code used during the analysis are provided to
the client at completion of the project to facilitate reuse or
reexamination of the work.