Software

Statistical Analysis

The statistical analysis may be driven by menus or by keywords, which are easy to operate and very flexible. The types of analyses, with their keywords, are shown below followed by the options that apply to them. Although experience has shown that the analyses are adequate for statistical assessment of data from numerous long-term studies. We are always pleased to consider including additional types of analyses or options to meet a client's requirements.

Type of analysis:

  • CHISQ 2xk chisquared analysis with optional stratification.
  • COUNT Simple frequency distributions.
  • EXACT Fisher exact test with optional exact test for trend.
  • HIST Histograms.
  • KWALLIS One-way analysis of variance by ranks with test for trend.
  • KWSTRATA Fry-Lee stratified version of one-way analysis of variance by ranks.
  • LIST Simple listings of data values, field names or files.
  • ONEWAY One-way analysis of variance, with Bartlett's test and trend test.
  • PEARSON Correlation coefficients for continuous data.
  • PETO Various 2xk analyses, from simple chisquared to full IARC analysis with stratification, taking into account whether the condition was visible in life, fatal or incidental. Both asymptotic and exact analyses are available.
  • PLOT Plots relationships between two variables. Can also show means, medians and fitted lines.
  • SPEARMAN Ranked correlation coefficients.
  • WEIBULL Maximum likelihood fit of Weibull distribution to data on time to visible tumour.

Options:

  • Validation: Check data by standard or user-defined criteria.
  • Transformation: Create new fields from existing data for analysis.
  • Select animals: Based on animal number lists or by more complex conditional statements based on values of specified fields.
  • Sex: Run reports for males only, females only, or both sexes.
  • Missing data: Various options avoid bias where pathology is incomplete.
  • Select organs: Analyse data for fields in selected organs.
  • Field lists: Define list of fields for analysis.
  • Tumours: Within a field list consider only tumours, only neo-plastic findings, or all fields. Optionally suppress multicentric tumours and/or metastases.
  • Cut-points: Analyse incidence above specified value or grade.
  • Sorting data: Output listings in user-defined order.
  • Significance: Present exact p values, p< format, or as codes. Optionally suppress output for any non-significant fields.
  • Confidence intervals: These are available in various analyses.
  • Types of comparison: When comparing groups, choose whether or not to include tests of pairwise differences (user-defined control group), trend (user-defined dose metameters) and overall heterogeneity.
  • Comparison groups: Users may choose which group or combination of groups is to be taken as the control group in comparisons. It is also possible to set particular groups to be ignored altogether. It is also possible to request all pair-wise comparisons to be made.
  • Factor: By default the analyses compare treatment groups, but comparison can be made of groups defined by any field.
  • Partition: If a field is defined as a partition, results are presented by level of the field and, where appropriate, combined over the partitions (stratified analysis).
  • Strata: Produces stratified analyses but only shows combined results.
  • Layout: Users can modify table headings, table and page numbers, page width and length, margins, tabs, underlining, decimal places, and other aspects of layout. Users can decide which of the output lines available for an analysis should appear. Users can add up to 6 lines of footnotes to their outputs.
  • Output: Results go to an output file and (optionally) to the screen. The output file may be printed, copied to a normal file or deleted at any stage. Plots are now produced in a form for printing directly to the laser printer.
  • Settings: Settings for an analysis are kept until altered. They may be stored and restored as required. An optional glossary describes the settings in operation for each output.
  • Session file: A record of an analysis session is kept.
  • Command file: A sequence of keywords to run analyses and to define options may be created directly or from a session file. These command files may be run in batch or interactive mode and can be used to define the complete sequence of analyses, thus facilitating rerunning following data modifications.