RSS Spreadsheet Forum, 27 March 1996

Association of Statistics Specialists Using Microsoft Excel

REPORT OF DISCUSSION

• good software to use with non-specialists
• has a strong user base in industry
• good for preparing ready-made templates for particular analyses
• dynamic linking - change the data and the output updates
• common environment with non-statistical users
• good range of functions
• Visual Basic in the background gives flexibility for add-ins
• ubiquitous - everybody has it
• excellent graphics for both teaching and using
• relatively cheap for individuals to purchase
• excellent value for institutions in terms of cost per hour of use
• flexibility in data handling - few restrictions
• allows students to get hands on data
• ability to give interactive demonstrations
• ability to attach notes to cells is useful (especially in Excel 7)
• continuous improvement
• durability - it will still be there in several years time

• moving average "trendline" is out of phase
• regression through the origin gives incorrect output
• x-axis labels on histogram wrongly positioned
• histogram cannot cope with unequal class intervals
• CONFIDENCE function uses z not t
• help on CONFIDENCE is totally wrong
• error bars are wrong in versions prior to 5.0c
• the function wizard cannot cope with array functions (e.g. linest)
• ranking does not handle ties properly
• there are errors in the inverse distribution functions
• RANDBETWEEN gives first value only half the frequency
• in two-sample t-test unequal variances, df are rounded not truncated

• consistency between functions (e.g. with respect to missing values)
• subscript/superscript tool on the Format toolbar
• a deleted cell should not be remembered in range operations
• an option to include/exclude hidden cells in calculations
• general facility to use non-contiguous ranges (e.g. multiple regression)
• ability to lock a data series on a chart to prevent dragging by mistake
• data analysis tools should dynamically link output with data
• better quality control - errors should be found before release
• Microsoft is too secretive - e.g. how does Solver really work?
• View Manager does not work on protected worksheets

• box and whisker plots
• proper histograms
• chi-squared test for contingency tables
• non-parametric test procedures
• a formal missing value indicator

Neville Hunt   02/04/96