University home » Faculty of Medical and Health Sciences » School of Population Health » SPAN » Manual

A Manual for SPAN

© Roger J. Marshall

Section of Epidemiology and Biostatistics, School of Population Health, The University of Auckland, New Zealand

Table of contents

1	Introduction
2	Installation and requirements
	2.1	SPAN on the Internet
	2.2	Version
3	SPAN methodology
	3.1	Partition representation
	3.2	Attributes
	3.3	SPAN Boolean  partitions
	3.4	Regular  partitions: positive attibutes
4	Windows SPAN
	4.1	Mouse input in windows
	4.2	Extracting contents of windows
	4.3	I'm thinking...
	4.4	The View function
5	Data preparation
	5.1	Free format
	5.2	Fixed format
	5.3	Excel spreadsheet input data
6	Main menu items
7	File and Edit
	7.1	File
	7.2	Edit .SPN and Edit file
	7.3	Edit:Split sample
	7.4	Edit:Select ...
8	Control file: Data input
	8.1	Control file
	8.2	Variables
	8.3	Format
	8.4	Grouped data
	8.5	Missing values and missing attributes
	8.6	Data restrictions
	8.7	Character Data
	8.8	Boolean combination lines
9	Control file: Creating attributes
	9.1	Attribute representation
	9.2	Interval and ordinal variables
		9.2.1   Range attributes
		9.2.2   Percentile and mean cutpoints
		9.2.3   Multiple cuts
	9.3	Nominal or binary variables
		9.3.1   Category representation
	9.4	Labels
	9.5	Direction of association
	9.6	Multiple attributes
	9.7	Using SPAN to edit or create a control file
	9.8	Missing value attributes
		9.8.1   Nominal variables
		9.8.2   Interval variables
	9.9	Special attributes
		9.9.1   The null attribute
		9.9.2   Index attribute
		9.9.3   Random attribute
		9.9.4   _all_ attribute
		9.9.5   The cross-validation specification
	9.10	Continuation indicator
10	Y menu
	10.1	OK
	10.2	Distribution
	10.3	Scatterplot /cross-tab
		10.3.1   Z: Adding a third variable. Rotating 3D plots		10.4	Transform
		10.4.1   Creating attributes from binary transformations
	10.5	Select Y
	10.6	By-group/cross-validate
11	Criteria
	11.1	Criterion
	11.2	Balancing parameter gamma
	11.3	Penalty parameter beta
	11.4	Complexity
	11.5	Priors for Entropy/Gini indices
	11.6	Costs for Entropy/Gini and Quality indices
12	Rank
	12.1	Attributes Rank Plot
	12.2	Fix positive attributes
		12.2.1   Multiple Y
	12.3	Effectiveness v Cuts plot
	12.4	ROC curves
	12.5	QROC curves
13	Strategy
	13.1	Constrain X space
	13.2	Force in attribute
	13.3	Iterative mode
	13.4	Tree
	13.5	Top m
	13.6	Decide positivity
	13.7	Set Extent parameters
14	Extent
	14.1	Size p_1 ....p_q
	14.2	Up to stated size
	14.3	A  and A-
	14.4	Number of float levels
15	Search
	15.1	The Search
	15.2	During the search
	15.3	The end of a search: complexity hull
	15.4	Penalty parameter setting
	15.5	Searching in iterative mode
	15.6	Search interrupts
	15.7	Search with by-group/ cross-validate
	15.8	Search in tree mode
16	Process
	16.1	Rectangle Diagrams
		16.1.1   Rectangles construction method
	16.2	Lists
	16.3	Distribution
	16.4	Statistics
		16.4.1   2x2 table statistics 16.4.2   Chi-square, generalised error etc 16.4.3   Risk matrix 16.4.4   Log-rank: incidence rates 16.4.5   Table of prior adjusted pseudo-counts
	16.5	Random
	16.6	Tree
	16.7	Test [training] sample
	16.8	Create added attribute
	16.9	Manual partitions [simplifying Boolean expressions]
17	Options
	17.1	Full Boolean reduction
	17.2	Attributes are primitive
	17.3	Tied partitions: use last found
	17.4	Missing attributes
	17.5	Detailed output
	17.6	Display graphics while searching
18	View and Window
	18.1	Output log
	18.2	Viewing data and .SPN files
	18.3	Added attributes
	18.4	Window
19	Help
	19.1	Website
	19.2	Topics
	19.3	Enter Registration #
	19.4	Manual
	19.5	FAQs
20	References

Appendices

Appendix 1	Lock and Key algorithm
Appendix 2	Size of search
	A 2.1	Number of partitions of size p1,...,pq
	A 2.2	Sub-search extent
Appendix 3	Partition criteria
	A 3.1	Within MSE
	A 3.2	Subgroup MSE
	A 3.3	Entropy
		A 3.3.1  Prior probabilities
	A 3.4	Quality index QI(r)
	A 3.5	Chi-square
	A 3.6	Odds ratio (Bayes)
	A 3.7	Log-rank
	A 3.8	Gini diversity
	A 3.9	Directional v. Non-directional measures
	A 3.10	Multiple Y measures
Appendix 4	Computational notes
	A 4.1	Timings and Turbo facility
	A 4.2	Program limits
	A 4.3	Error messages
	A 4.4	Bugs
	A 4.5	Limitations
Appendix 5	Manipulating Boolean formulae
	A 5.1	Reduction rules
	A 5.2	Transposing partitions
	A 5.3	Expanding out partitions
	A 5.4	Variable dependent Boolean reductions
	A 5.5	Null partitions
Appendix 6	Skipped partitions
Appendix 7	Assessing statistical significance
	A 7.1	Non-randomness of generated partitions
	A 7.2	Significance of the best partition