CROSSTAB Vers. 1b2 (c) 1996 by Ralf Gesellensetter *** A scientific programme for statistical processing of one sample data *** SPECIFICATIONS - Read Table as Comma Seperated Values (CSV) - max. 255 fields, 10000 lines - Automatical calculation of common statistics (mean, StdDev, median etc.) - Discrete and classified frequency tables (incl. Kolmogoroff/Xý-Fit-Test) - Crosstables, automatically (by now for integer fields only) - Automatical calculation of 2-dim statistics (covariance, correlation) - Test correlation: t-test (NormDist), Spearman-Pearson Ranksum(ParFree); independence: chi^2 and Woolf's G-test with Yates' continuity correction (*) for 1 df (if min E(Nij)<=30) plus Fisher's Exact Test for thin 2x2 tables - Calculation of a chart of diverse given random variables - Compute prob's and quantiles of normal, t, chi square and F-distribution - NEW: Log-File, Batch-Mode, Browser; Universal output selection COPYRIGHT This beta version may be copied freely. Frequent users must register, though (see REGISTER.TXT). Any commercial abuse or sale prohibited. Contact 'n' Credits through... email: gesellen@namu26.gwdg.de Web: http://namu19.gwdg.de/gesellen/ MANUAL - some explanations ========================== Menu FILE - Browse Browse an ASCII File (like OUT- or LOG-File) using arrow keys. Quite simple and slow, but without any size limitation. - Execute Provides sufficient memory. No guarantee for further stable continuation. Menu TABLE - Open Tables must be according to the CSV standard: Each line makes a data row; the field's contents are seperated by commata. Text variables should be quoted. Comments are allowed in lines starting with '#' etc. The first line is interpreted as titles in the case that it containes no purely numeric data at all. - Create This is basically to test the tests: A table of about 10 random fields is created. The amount of entries should't exceed some hundreds as the calculation takes its time (especially chi square random vars). - Save Write CSV file or join existing data... Appending a table to an existing file involves omitting its headings. It doesn't make sens in most cases, especially if the amount of fields varies. Statistics - Univariate - Standard Should always be applied first. Gives some information on the fields you select. Like all the other statistic functions, it runs automatically after you have once choosen the very fields (unless the output channel is set to monitor, see below). - Freq After choosing the fields of question, frequency tables will be run: a) discrete fields (int) Any appearing value is listed with its frequency (absolute, cumulated and relative). Then (if OPTIONS - 'Apply Tests' is tagged), the Kolmogoroff normal fit test is applied (fixed niveau: 5 p.c.; result: parfree or normal). [As for the tests, see below] b) continous fields (flt) Frequencies are shown classified (about sqrt(n) classes). Expected cases are calculated for the chi square fit test (normal distribution). BIVARIATE - Covariance Calculates mutually for all selected fields: - empirical coefficient of correlation rho - Probability for a more extreme result at rho=0 (2-tail) Tests - Pearson-Spearman Rank-Difference-Square-Sum Test ( coefficient of midrank correlation) - Crosstabs Works still only for discrete data (int). Apart from a cross table there is the option of the chi square test of significance. Its validity provides at least 30 observations and expected frequencies of at least 5 in each cell. Therefore there is a coninuity correction of .5 in the case of fourfold (2x2) tables with thin cells (according to Yates). In the latter case there is also done Fisher's Exact Test (hypergeometrical) and Woolf's G-Test (approximative/ logarithmic/ Likelihood Quotient). Version 1b2 provides Pearson's corrected contingency coefficient, too. OPTIONS - Output: Channel and page format - System: some settings, accurancy of output (no internal effects) - Apply Tests (switch) - Log File: if tagged, all called functions are documented in CROSSTAB.LOG (see below). EXTRA - Calculate Compute values of the most common distributions and their inverses. - Batch-Mode A batch file may be run. For instance you can use parts of a logfile. Thus a great bunch of work can be done automatically... ----------------------------------------------------------------------------- USAGE - Special Features: - Provided there is any data, you may scroll the table by means of the arrow keys. - While selecting the fields on which you want to do statistics, clicking on the 'Selected' counter deselects all entries. - In order to reset the Log file (always CROSSTAB.LOG) click on the tagged menu entry 'Log-File' while pressing Shift. - If you respond to a 'KEY!' demand (or similair requests) by hitting ESC, you won't need to respond to any further form feeds during the current function. - Generally, running functions may be aborted by pressing ESCAPE. Don't be too wary about that: you'll be asked for confirmation. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Warnings/Possible Problems - There is no alarm of low memory - you'll know when you meet some bombs... - I don't know what happens if you try to send your output to a dead channel (e.g. a printer being off-line)... - The approximation of the F-distribution is quite poor for less than 10 df. (any proposals for better formulas?) - The inverse Chi Square function is iterative and thus quite slow. ****************************************************************************** Possible Extensions/Upgrades Extensions will naturally depend on feedback, if any. There is still better applications like sas or spss, so maybe there is not any deman at all for this application. In that case I won't bother to spend any more night on it... Otherwise I am thinking of... - some graphics - more tests - improvement of the approximations - even more columns - import/export more different formats - more tables/samples at once - implementation of a parser/command line interpreter, possibly based on fields. - even more automatism by interpreting written LOG files... ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^