EpiData Analysis
Copyright 2002-2006 Epidata Association www.epidata.dk
Issues in testing or development
- SPC graph options extended - exclude observations from calculations:
/exp=x (x: observation number, e.g. /exp=18)
/exv=x (x: observations with Y variable larger than value excluded) - ! - command prompt adapted. WIthout option /Open the command prompt will close again
/open (keep command prompt open - useful for checking errors and parsing of commands)
Latest public release - Information below
Version: Jan 19th. 2006
Known issues to extend or include in coming versions
- Odds Ratios and RR in tables with Rare data (0 in cells)
- Correlate command results in "floating point error" when correlate cannot be calculated
- Contents of strings should be converted to UPPER in some logical statements to work. e.g. count if upper(v) = "TEST"
- A new principle replacing setting format will be devloped
- Tests of homogeneity (interaction) in summary stratified 2x2 tables not implemented
- Means and Kwallis do not allow string variables for the group variable
- Percentages in pie charts are not always correct.
Latest build
Version 1.1.2 build 62
- Principle of preventing running twice reversed
Now: file preventdouble.ea installed on all winXP+ versions
Users can delete this to run several instances
Before: file rundouble.ini should be created on all versions
Before latest
Version 1.1.1 at build 62
- Help file shown in separate window
- Press F1 to show commands help file in separate window
- Show extended command help (if made):
write command in prompt and press F1
if not found that command is shown in commands.htm reference - Show html file in help window: help xxx
- Modified Output file principle
- Output file turned on at initiation of programme EAoutput.htm
- Saveoutput - same as logopen
- CloseOutput - same as logclose
- "Set stylesheet external=on[off]"
when off a copy of the stylesheet is included in the output html file.
when on: a link to the stylesheet file is included instead
- User Control of more than one concurrent instance of Analysis
- Modified Commands
- temporary select "if" must be placed at the end of a command
except: select and count "if" must follow select/count
- temporary select "if" must be placed at the end of a command
- New commands
- aggregate (abbreviation agg) - create aggregated statistics (collapsing data)
- stattables (abbreviation stattab) - tabulate aggregated statistics
- label variables (label
"text" ) - labeldata "text" (label the data file)
- Extended Commands
- Graphs
- SPC charts: Test1 modified for Run Charts
- SPC charts: Missing in count variable allowed
- SPC graphs - tests are shown in a table
- Graphs addition of Y axis text: /Ytext=
- Graphs addition of X axis text: /Xtext= (default is variable label of x-variable)
- Default colouring of bar charts: Each bar separate colour
- "Set graph colour" defines bar chart colouring (default value: 123456)
- "Set graph filename show" Add name of graph file below the graph - default = off
- "Set graph filename folder" Add complete folder for graph file to html output - default = off
- Variable label addition to tables, graphs etc. controlled with "set var label" and "set var name"
- ERASEPNG - new option added: /ALL (will include all png files, without only graph*.png are included)
- Frequency tables
- Variable label added
- Cumulative percentage
- Options D0 D1 (zero or one decimal on percentage)
- NC - no cumulative percentage
- F8 now activates file indicated in "Set start page"
- Link to introduction pdf file added to help menu
- Right mouse click on variable (F3 window) shows all variable definitions
- "Results" sufficient for "var results"
- "results clear" sufficient for "var temp clear"
- Graphs
- Modifications to set
- "Set window font size" defines size of "F3/F2/F7" windows on right side of screen
- "Set editor font size" defines size of font in editor
- "Set Command History" changed to "Set History Command add" (add pgm commands to history)
- "set HISTORY COMMAND PGM" default is now OFF -
on: add all commands in a pgm to history - "set RECODE INTERVAL TEXT" the string inserted with "recode to", default "-"
- "set Generate type" changed to "Var generate type" (Default type for variable generation)
- Bugs fixed:
- Copy and Paste from output window to editor or command prompt was not working
- Various small formatting or parsing bugs (e.g. copyfile with spaces in filenames)
- error with parsing of if command - which could replace data
- Means group error with one observation in a group
Version 1.0 at build 43
- Added or modified
- Release number removed - only version and build shown from now on
- Possible to run Exe file without any other file (Graphs will not work)
- Bugs fixed:
- Define in CHK files are accepted
- Conditions for test 2 in SPC charts adapted
- Empty lines at top of text type logfiles removed
- TAB in pasted lines into editor changed to spaces
Version 1.0 release 0 build 40
- Added or modified
- Refined style sheets for output
- Reference to observations must be in []
- Observation (record number) can be referenced with _n. E.g. age[_n -1]
- Introduction document added to installation
- Error given if group variable in means or kwallis is of type string
- Bugs fixed:
- Certain if statements not parsed correctly
- Append with records marked for deletion.
Version 0.9 release 6 build 39
- Added or modified
- Refined style sheets for output
- Bugs fixed:
- Marginal percentages with etab command.
- Append with records marked for deletion.
- Headers for "first word" rec files could be wrong in some instances
Version 0.9 release 6 build 38
- Added or modified
- Tables will have collapsed borders
- Chi2 are ALL uncorrected values now (except for MH summary)
- Bugs fixed:
- Chi2 values for very large numbers could be wrong
- Append with records marked for deletion.
Version 0.9 release 5 build 37
- Added or modified
- Documentation revised and clarified (output definition)
- Added to help menu: Edit setup file (epidatastat.ini)
Version 0.9 release 5 build 36
- Added or modified
- A "select font" on first startup to guide the user for efficient setup of screen
- Improved documentation of output formation
- Table command can show categories defined as missing
- New option controls if more than one instance of the programme can run
- Aspects of handling missing value further implemented.
- Parsing of strings changed to exclude trailing blanks: e.g "Lion " is the same as "Lion", but not the same as " Lion".
- sorting of tables corrected
- Tests of homogeneity (interaction) in summary stratified 2x2 tables removed until final stable solution is found
- Bugs fixed:
- "Pointer error" in tables or describe fixed
- Error in table estimates for 2x2 tables
- Recnumber now returns number within current select
- set listrec was not working
Version 0.9 release 5 build 28-35
Not available for public testing. Different strategies to solve bugs were tested.Version 0.9 release 5 build 27
(This is a minor release.)- Added or modified
- Version now also reads a comment from http://www.epidata.dk
- Bugs fixed:
- Emin in Tables also checked marginal totals
Version 0.9 release 5 build 26
- Added or modified
- This is a major release. Release number introduced again. Builds will indicate smaller changes
- Reading of data files and chk files final (see note on recnumber definition above)
- Two table commands are now working: Tables (for general tabulation) and Etables (Epidemiology tables)
- Several Output class definitions in one line allowed (separate with ,)
- Improved command history window (F7)
- Notice: Graphs showing as ! in output window is not a bug, but due to shift in folder from where graph file is.
- comparing a string variable to a constant will remove trailing blanks, e.g. "test of" = "test of "
- Bug removal:
- Table command - thorough testing of correctness of estimates and inverting tables.
- Table stratified - estimates for subtables fixed. Command changed to etab
Faulty calculation of summarised RR and OR, plus tests of interaction - fixed - Processor dependent parsing of count and if (serious bug, difficult to catch and solve)
- Boolean variables corrected (Y N and . read correctly)
Version 0.9 build 23-25
- Internal builds to find solutions for serious bugs on some PC's
Version 0.9 build 22
- Added or modified
- Reading of data files and chk files final (see note on recnumber definition above)
- Output adapted to W3c (www.w3c.org) standards, see pdf file in "docs" subfolder.
- Simplification of control of what is shown (commands, info, results): set show definitions
Note some inconsistencies are still there - set parameter for table design added
- Reading of files with encrypted fields (prompt for password or add /key)
- Major commands can be used with first three letters
- Reading of missingvalue definitions
- Output changed to tables (improves speed)
- Several interface adaptations
- Added command history window (F7)
- Much improved examples on opening pages
- Improved reporting on append
- Bug removal:
- Crash when selecting block in output fixed.
Version 0.9 build 21
- Added or modified
- Added row column total percent with special formatting of each
- Enhanced find function
- Command Prompt can be invisible (but exists)
- Faster output for variable and directory lists
- Max cases for Correlate and Regression increased to 5000
- command CORRELATE correctly rejects when number of observations > 5000
- Various other refinements of interface
Version 0.9 build 20
- Added or modified
- set parameter for save pgm file changed to "Set history name="
- Illegal national characters in field names replaced on reading
- Improved feedback with append files (when structure is different)
- Bug removal:
- Option Quiet in describe reversed /q=show lacking: hide output.
Version 0.9 build 19
- Added or modified
- set parameters unified (NOTE MANY CHANGES)
- note that to invert 2x2 tables use "set table epi=on"
- output and selection of estimates from tables command
- Bug removal and modification:
- table command labels and inversion incorrect in b18 (fixed)
Version 0.9 build 18
- New
- search in output window and help files
- APPEND command: add data from other files of same structure
- option /quiet or /q added to means and describe
- Reading data can be stopped by ESC
- PgUp and PgDn moves output up and down.
- Several options to control table output (rr or percents)
- Tables with 2x2 data are inverted if OR or RR requested
- Added or modified
- headers corrected when writing data files
- testing routines finetuned (try "runtest validate")
- Internal representation of float variables finalised
- Negative values in recode possible
- Assert and Assert error internal variables not reset after means,describe and count
- Deleted Record Status obeyed when reading data (set readdeleted=off|on)
- Bug removal and modification:
- Blank output screen after cls fixed
Version 0.9 build 17
For few hours a prebuild 17 could not show graphs- Added or modified
- Internal representation of float variable enhanced
- Table output adapted (Odds Ratio's)
- Bug removal and modification:
- Creation of boolean fields fixed
- Negative values in result variables fixed for mean
Version 0.9 build 16
- Added or modified
- update shows which values are changed with if statements.
- if started with a pgm file, that file will be executed.
- internal representation of float variable changed from double to extended
- Testsystem improved (e.g. try startnist.pgm)
- Bug removal and modification:
- String variables of length 1 are saved correctly
- negative values in result variables fixed
- lre() function corrected for negative values
- Calculation of 10,25,75 and 90 percentiles corrected
- recode function corrected to include limits btw. groups
- Various interface problems
- Documentation
- Adapted Commands.htm and Howto.htm file.
Version 0.9 build 15
- Added or modified
- Graph button on workprocess toolbar working again (bug introduced in build 14).
- Number on button 4 removed (lacked removal in build 14)
Version 0.9 build 14
- Added or modified
- Warning on Chi value imprecision for N>950 in kwallis
- Adapted work process toolbar and menus
- bar graphs now show value labels at bottom
- runtest command gives table of results
- set allstats changed to allstatistics
- Bug removal and modification:
- Parsing of strings "a "+"b"="a b"
- Variable On parsing several variables to freq or describe "a no data" error could should up if one variable was all missing.
- Variable lists extended (e.g. freq a-d)
- Confidence intervals: mean+/- 1.96*sqrt(Variance/N) (was not always correct)
- Median corrected.
- Documentation
- Adapted Commands.htm and Howto.htm file.
Version 0.9 build 13
- Added
- set: set Viewer Font Charset= (to accomodate Chinese characters)
- User configurable window placements (saved in epidatascreen.ini)
- New command: eraseallpng /NOCONFIRM (will erase all graph????????????.png files)
- Graph font size specification: set graphfontsize= (default pitch 10)
- Bug removal and modification:
- Access violation when creating graphs
- On parsing several variables to freq or describe "a no data" error could should up if one variable was all missing.
- Browser now shows only selected records
- Various smaller bugs or adaptations
- Documentation
- Adapted Commands.htm and Howto.htm file.
Version 0.9 build 12
- Added: /append and /close parameters for logfiles
- Bug removal:
- Incorrect autosaving of pgm files (did not read parameter)
- Documentation
- Adapted Commands.htm file.
Version 0.9 build 11
- Added set: set labelshow= on/off/both (shows valuelabels, values or both in tables)
- "How to.." and "commands" added to help menu
- Adaptation:
- Added variable names and labels in tables
- Bug removal:
- Incorrect values with describe for median and percentiles
- Frequency tables incorrect with several variables
- Incorrect return of select prompting: select age < (?what age?)
- access violation if trying to define variables and no data file open
- defining dates are now dates e.g.: define d
- Documentation
- Adapted Commands.htm file.
Version 0.9 build 10
- Added close option to read, e.g. read "testdata\bromar.rec" /close
- Adaptation:
- 45 degree x labels on SPC graphs as default
- Tickmarks on all graphs except with /NOXTICK /NOYTICK
- A pre build 10 existed for a few hours without these two:
1 Replacing datafile with savedata controlled by set replacedatafile
2 Replacing graph file controlled by set graphreplacefile
- Bug removal:
- Regress now activated by workprocess toolbar. (was desribe)
- Naming of autosaved graph files changed (could contain :)
- Minor interface bugs - e.g. adding of double click to F2 window
- Main Menu file submenu now acts with close
- Documentation
- Adapted Commands.htm file. Introduced howto.htm with suggestions.
Version 0.9 build 9
- Fixed bug which showed Field Type wrong in F3 window for byte (string 1) variables
- Fixed bug when using byte variables in describe command
- Adapted info.htm opening file
Version 0.9 build 8
- Graph dialog finalised (still lacking details of axis control)
- Further changes of reading - recognicing formats - in delimited data files
- Fixing bugs in relation to missing values in frequency tables
- Adaptation - information - to users when reading errors in data
- Adapted info.htm opening file
Version 0.9 build 7
- Graph dialog introduced
- Changes of reading - recognicing formats - in delimited data files
Status at version 0.9 build 1 to build 6
- Complete restructuring of internal structure in program to make it stringent in style and separate interface handling from commands.
- Statistical process control graphs and principles in collaboration with Gruk Norway
- Interface policy finalised - menus for beginners, commands for experienced users.
- Finalisation tests commenced
Status at version 0.9.0
- Unfortunately Salah Mahmud had to stop participation in the development due to other committments in autumn 2003. The core parser made by Salah Mahmud constitutes a key part of the project. Therefore the parser is referenced as a separate entity. Salah also contributed to parts of the early documentation.
Early development
Versions prior to 0.9
- New commands
- Loop command, to repeatdly execute a command or a command block
- View command, to display HTML files in the output window.
- New functions/system variables
- Sysdir: system variable, returns the path to the folder containing EpiData Analysis (EpiDataStat.exe
- The new Program editor is functional
- The output window was enhanced to allow copying HTML to clipboard and to support moving between viewed documents like an Internet browser
- improvement in parsing commands
- Reading/writing data
- Epianalysis can now read all the new data types introduced in Epidata 2.2 (encrypted fields are ignored)
- Optimization
- Reading REC file is slightly faster now
- Tabulating one-char string variables (eg sex=[M,F]) is as fast as tabulating boolean or integer variables