Other statistical tests included in ABstat are confidence intervals, sample size estimates and probability calculations.
See Feature Summary for a complete list of commands.
Your data can include character variables as well as numeric. Character data are especially useful as grouping variables.
For instance, if you have a Weight variable (numeric) and a Sex variable (character), you can easily do an independent t-test comparing the weights of males versus females.
And rather than cryptic codes of M and F showing up in your reports, you can easily do value labeling so even though the codes might be used in the data, the reports show Male and Female.
Survey questions can be open-ended (verbatim responses), numeric, multiple choice or all-that-apply, providing all the basic question types used in surveys.
All-that-apply questions, where the respondent can choose multiple answers instead of just one, are as easy to set up and work with as single-response questions.
Multiple choice and all-that-apply questions utilize simple choice tables for defining valid choices.
If your survey uses an identical scale (such as strongly agree to strongly disagree) for several questions, you need only define that scale once, simplifying survey setup.
And you have great flexibility in defining the choices.
Choice codes (what you actually enter in the data) can be as short as 1 character (typical) or as long as 8 characters so you could enter meaningful words or abbreviations instead of just a,b,c.
Any choice table can have up to 128 entries so state tables or even larger tables can be used.
And once you've created choice tables you like to use, it's easy to incorporate them into future surveys because you can create your own master file of choice tables and easily copy choice tables from one file to another.
There are two primary reports used for survey analysis: the Count report and the Banner-and-Stub cross-tabulation report. Both reports work with all-that-apply questions or single-response questions, adjusting percentages as required.
The Banner-and-Stub report goes the Count report one step further. It might also show how people liked a new product, but in addition, it shows how responses vary by sex or age or other demographics.
Here's an example:
But not all surveys require paper. The On-Screen add-on product allows you to administer the survey directly on the computer without paper forms.
This is great for in-store, trade show, shopping mall, or employee surveys. And since it allows "skip patterns", it is also quite useful for telephone surveys or other interviewing applications.
The On-Screen program is self-contained and designed to be used by someone who knows absolutely nothing about the program as all instructions are right on the screen.
Each question is automatically formatted onto the screen with the entire question wording at the top and the possible answers (for multiple-choice and all-that-apply questions) appearing below.
With Windows, a simple mouse click is all that's required to answer most questions.
"Skip patterns" provide for tailoring each questionnaire to each respondent by asking only relevant questions based on answers to previous questions.
Example: you may have a question which asks for a person's sex followed by two completely different sets of questions depending upon their gender.
On a paper survey form, instructions would have to guide the respondent to the next appropriate question.
With On-Screen, the respondent automatically gets asked the right questions.
ABstat feature summary and program specifications
Statistics
ANOVA: Analysis of variance allows up to 4 way analysis with unequal-n, zero, one, or all factors repeated, and several post-hoc tests and means report.
Chifit: Chi-square goodness of fit test.
Confidence intervals around means or proportions: selectable degrees of confidence.
Correlation matrix: Pearson's r correlation and probability.
Simple Cross-tabulation: with Chi-square and Fisher's exact test (if 2x2 table).
Descriptive Stats: mean, standard deviation, variance, standard error of mean, coefficient of variation, minimum, maximum, range, total, median, mode, skewness, kurtosis.
Frequency report: values, frequencies, percents, percentiles, z-scores, histogram.
Kolmogorov-Smirnov test for uniform distributions.
Kruskal-Wallis: 1-way ANOVA for ranked data.
Mann-Whitney U test.
Probability calculator: for Chi-Square, F, Student's t, binomial, Poison, Z-score distributions.
Regression: simple and multiple linear regressions, up to 30 independent variables. Forward and backward stepwise, estimate and residual calculations, Durbin-Watson. Used with transform for non-linear regressions.
1-sample runs test.
Wald-Wolfowitz: 2-sample runs test.
2-sample Sign test.
Spearman rank: correlation matrix.
Sample Size Estimates: for means or proportions; selectable degrees of confidence.
Independent t test, Paired t-test and Population t-test.
Wilcoxon: matched pairs signed ranks test.
Transformations
Transform: Modify variables using any algebraic expression, including variables in your data file. Put values in character variables. Specify any subset of records to modify. Over 35 algebraic and logical functions available, including use of parentheses, and's and or's.
Build: Modify a variable using an expression involving data in a different file.
Pull: Add variables from a different file.
Append: Add records from a different file.
Sort: Sort your file by up to 20 variables.
Rank: Create rank-order variables.
Random Numbers: You can generate a random variable with desired ranges of values in either uniform or Gaussian distributions.
Recode: Create or change coding schemes. Convert ranges of values into multiple-choices.
Date and Time transformations: Do arithmetic and format conversions with variables which contain dates and times (in various formats).
Graphics
High-resolution: Analysis grade graphics for histograms, bar-charts with error bands (simple, side-by-side, or stacked), pie charts, scatter plot (with optional regression line and equation), line graph, multi-line graph. Scaleable scatter, line and bar graphs. Graphs are made directly from raw data without having to summarize values into special variables.
Box-and-whisker plot and the histograms in the Count and Frequency reports provide fast, low resolution graphics for quick analysis.
Survey analysis
Choice Tables: Define value labels for one or more variables. Use to ensure data entry accuracy and report categorical variables. Create new choice tables or re-use from existing files. Can be weighted to generate means and related statistics.
Count Report: Reports multiple-choice and all-that-apply questions indicating the frequencies and percents of how each question was answered for all your respondents or any subset.
Banner Report: The banner and stub cross-tabulation report is the most powerful feature for survey analysis. You have over 20 options available, including percent selections, case weighting and statistics. You can easily define your most used "normal" report options to make your analysis easy. Formatting and pagination are automatic (up to 40 "banner points" per page). All you have to do is select the questions and options you want on the report.
Other features
Select a subset: Use this versatile feature with virtually any other function to enable subset analysis. Specify your subset and then edit, list, or use any statistical or reporting command.
Breakdown: An automatic "select each value and repeat the following command" function provides easy repetition of any command for all values in a variable.
Case Weights: An integer case weight variable can be specified. This is especially useful when your "raw" data is already summarized. Simply specify the number of times each record is to be counted in the case weight variable. Non-integer case weighting is also available in the "Count" and "Banner" commands.
Options Setup: A variety of user options can be selected with this command to help customize the program.
Log On/Off: A log file can be produced of any commands during your session. This allows you to create a history of all commands performed and to edit and rerun those same commands later as a batch file. Can be considered a "macro record" function.
Batch: Any commands you do "live" can also be recorded in a batch file for automated analysis/reporting at a later time.
Especially useful when you do repeat analysis on data accumulated over time.
This function also allows "parameter substitution," providing great flexibility.
Can be put in single-step mode for step-by-step operator control or modification of commands.
Can be considered a "macro play" function or almost a programming language.
Missing values: Full missing value support is provided for each statistical command. Five missing value treatment types are available with the default for each command set to maximize the amount of data available to that command.
Cancel: You can cancel any command while in progress. If you cancel a batch process, you can resume the batch process (perhaps skipping over some commands).
Help makes extensive use of definitions and cross-references and has a very complete index. The context sensitive Help button, for virtually every dialog box in Windows, explains each of the command's options. In DOS, F1 or "?" gets you to the help screen and F2 gives access to variable and file names as needed.
Define Headings: You have complete control over your headings with options for centering or right-justifying text on heading lines and choosing whether you want the headings at the top or bottom (or both) of each page.
Data List: This shows raw data or descriptions (value labels). You can select the variables (questions) and even the records you want listed.
Structure List: This shows how your file is defined in terms of the variable (question) names, how they are defined (character, numeric, etc.) and full question wording. You can edit your structure at any time - even after your file has data.
Type a File: Include information from a text file in your report output with this command.
Remarks: This command enables you to add comments directly from your keyboard to your report output.
Program Limitations
Supports files with up to 512 variables (questions) and up to 32,000 records. Record length can be up to 6143 characters.
Variables can be numeric (either ASCII or floating point) or character. Character variables can be from 1 to 255 characters in length (for verbatim responses).
Multiple-choice and all-that-apply variables can have codes from 1 to 8 characters in length and descriptions (value labels) up to 30 characters.
Any choice table can have up to 128 choices. Variable names can be up to 10 characters and variable descriptions (question wording) can be up to 245 characters long.
Complete interoperability between DOS and Windows versions.
Import/Export Capabilities
To/from ASCII (comma or tab delimited or fixed format), dBASE, Lotus, Excel, and Foxpro.
System requirements
IBM compatible computer
Windows 3.1 or later for Windows version
DOS 2.0 or later (3.1 or later for network versions)
80386 or better processor for Windows version
Memory and disk requirements are so small it will run on virtually any machine
VGA graphics for Windows version
a mouse for Windows version
An Intel math co-processor speeds processing but is not required
Network versions, enable file-sharing, record locking and password protection. Works with virtually any network.
