Play and Learn 2

home *** CD-ROM | disk | FTP | other *** search

/ Play and Learn 2 / 19941.ZIP / 19941 / EDUCMATH / STATHELP / SHZIPARC.EXE / WATSTAT.HLP < prev next >

Wrap

Text File | 1991-12-26 | 16.3 KB | 257 lines

*DATA HELP Type the Number of your choice & Press the [ENTER] key. Press [ESC] to remove a Choice Box without registering a choice. ESCaping from the first Box will return you to the MAIN MENU; ESCaping from any other will return you to the previous Choice Box and allow you to select a different option. *RANDOMNESS Type the Number of your choice & Press the [ENTER] key. Choose RANDOM SAMPLING if a definite random sampling scheme was used to draw the sample from a pre-defined population. The sampling scheme could be a SIMPLE RANDOM SAMPLE, a SYSTEMATIC RANDOM SAMPLE, or a COMPLEX PROBABILITY SAMPLE, such as a stratified random sample or a multi-stage cluster sample. Choose RANDOM ASSIGNMENT if cases were randomly divided into sub-groups, as in a true experimental design or randomized clinical trial. (If both random sampling and random assignment were involved - a rare event - either choice '1' or '2' may be selected at your option). Choose NON-RANDOM SAMPLING if no random process was used to generate the sample and neither of the above options applies. This choice would be appropriate if cases represent a CONVENIENCE SAMPLE, AVAILABILITY SAMPLE, QUOTA SAMPLE, or a whole POPULATION. It is also appropriate for QUASI- EXPERIMENTAL designs, where cases are divided into naturally occurring groups or assigned to treatments according to judgmental criteria. *INDEPENDENCE Choose OPTION 1 if BOTH of the following conditions hold: a) The analysis will divide the cases into sub-samples for comparison, but there will be NO MATCHING of cases across sub-samples AND there was no matching inherent in the procedure used to draw the sample. b) Each case will be represented ONLY ONCE in the analysis: no before-after or repeated measures will be involved. Choose OPTION 2 if the analysis will compare MATCHED SUB-SAMPLES, e.g., husbands vs. wives or people with & without a certain disease who are matched on age and sex. Choose OPTION 2 if each case will be represented more than once in the analysis, as in a before-after, repeated measures, or cross-over experiment. Choose OPTION 3 if the sample will not be split into sub-samples and no between-group comparisons will be made. *QUESTION Type the Number of your choice & Press the [ENTER] key. Choose OPTION 1 if you're interested in describing the distribution of only one variable AND no comparisons will be made between sub-samples. Choose OPTION 2 if you're interested in describing the association or correlation between two or more variables AND no comparisons will be made between sub-samples. Choose OPTION 3 if you're interested in describing or measuring differences across two or more sub-samples. Choose OPTION 4 if you're interested in sub-sample differences (as in '3'), but you also want to control or adjust for some extraneous variable(s) that you have measured for the same cases. *COMPLEX Type the Number of your choice & Press the [ENTER] key. Choose OPTION 1 if you have only one variable to describe or summarize AND no sub-sample comparisons are involved. [You MUST choose this option if you chose "Univariate Analysis" in the previous Choice Box.] Choose OPTION 2 if exactly two variables will be used in the analysis. Note that each sub-sample breakdown counts as a variable. Choose OPTION 3 if three or more variables will be used in the analysis AND only one of these will be designated as the DEPENDENT variable to be explained or predicted by the other (independent) variables. Choose OPTION 4 if multiple variables will be used in the analysis AND two or more of these will be designated as DEPENDENT variables to be explained or predicted by the other (independent) variables. Choose OPTION 5 if multiple variables will be used in the analysis BUT no variable(s) will be designated as DEPENDENT. WATSTAT will assume that all variables are INDEPENDENT variables. [ NOTE: if you choose OPTION 5 here, you should also choose "Not Applicable" on the next Choice Box, which asks for more information about your DEPENDENT variable.] *DEP VAR. [Levels of Measurement are defined near the end of this message in a section labelled BASIC TERMS & CONCEPTS. Scroll to that section if you need a review of terminology.] Choose OPTION 1 if the dependent variable(s) is measured on a NOMINAL scale AND if it would not be appropriate to dichotomize it. Choose OPTION 2 if the dependent variable(s) is in the form of RANKS or if it is measured on an ORDINAL scale that could be transformed to ranks. This option also assumes relatively few ties: as a rule of thumb, fewer than half the cases should be tied on the dependent variable. Choose OPTION 3 if the dependent variable(s) is a set of 3 or more ORDINAL categories, so that all cases in a given category are tied. Choose this option also if most but not all cases are tied. Choose OPTION 4 if the dependent variable(s) is measured on an INTERVAL scale that produces either a continuous distribution or a distribution of 3 or more INTERVAL-LEVEL categories. Choose OPTION 5 if the dependent variable(s) is DICHOTOMOUS, irrespective of its level of measurement. Choose OPTION 6, "Not Applicable", if no variable will be designated as a dependent or "outcome" variable. You will be asked to specify levels of measurement for your variables in the next Choice Box. ---------------------- BASIC TERMS & CONCEPTS ---------------------------- Variables measured at the NOMINAL LEVEL identify differences between cases, but assume no underlying hierarchy along which cases can be ordered from "lowest" to "highest". Examples of nominal variables: race and hair color. (The special case of Dichotomous Nominal Variables is noted below.) Variables measured at the ORDINAL LEVEL identify differences between cases in such a way that they can be ordered from "lowest" to "highest", but do not specify how much lower or higher any case is relative to any other. Examples of ordinal variables: class rank, tennis seeds, and the order in which horses finish a race. A "pure" ordinal variable distinguishes each case from every other so they can be rank-ordered without any ties. In practice, ties commonly occur, and statisticians have devised ways to compensate for them. If there are relatively few ties simple adjustments to rank-order statistics usually suffice, but when there many ties an alternative to rank-order procedures may be needed. Such alternatives are commonly used for "partially ordered" scales, which arrange cases into a hierarchy of categories but make no distinctions within categories. With this sort of variable, every case is tied with at least one other case. Variables measured at the INTERVAL LEVEL identify differences between cases in such a way that they are not only ordered from lowest to highest but the SIZE of their differences can be stated in terms of a UNIT OF MEASUREMENT. Examples: Age measured in years and Income measured in dollars. DICHOTOMIES: If a variable divides cases into only two categories, it is often legitimate to assign arbitrary scores to its categories, e.g., 0 & 1, and treat it as an Interval Variable even if it was initially defined as a Nominal or Ordinal Variable. *IND VAR. [Levels of Measurement are defined near the end of this message in a section labelled BASIC TERMS & CONCEPTS. Scroll to that section if you need a review of terminology.] Choose OPTION 1 if ALL independent variables are Nominal and it is not appropriate to dichotomize all of them. Choose OPTION 2 if ALL independent variables are in the form of RANKS or if they are measured on ORDINAL scales that could be transformed to ranks. This option also assumes relatively few ties: as a rule of thumb, fewer than half the cases should be tied on any independent variable. Choose OPTION 3 if ALL the independent variables are sets of 3 or more ORDINAL categories, so that all cases in a given category are tied. Choose this option also if, for all the independent variables, over half of the cases are tied. Choose OPTION 4 if ALL the independent variables are measured on INTERVAL scales. Choose this option also if some or all the independent variables are DICHOTOMIES, even if the dichotomies result from NOMINAL or ORDINAL scales. Choose OPTION 5 if the independent variables consist of both NOMINAL and ORDINAL scales. This would be appropriate if each of the variables would fit under OPTIONS 1, 2, OR 3. Choose OPTION 6 if some independent variables would fit under OPTION 4, but others would best fit under OPTIONS 1, 2, OR 3. ---------------------- BASIC TERMS & CONCEPTS ---------------------------- Variables measured at the NOMINAL LEVEL identify differences between cases, but assume no underlying hierarchy along which cases can be ordered from "lowest" to "highest". Examples of nominal variables: race and hair color. (The special case of Dichotomous Nominal Variables is noted below.) Variables measured at the ORDINAL LEVEL identify differences between cases in such a way that they can be ordered from "lowest" to "highest", but do not specify how much lower or higher any case is relative to any other. Examples of ordinal variables: class rank, tennis seeds, and the order in which horses finish a race. A "pure" ordinal variable distinguishes each case from every other so they can be rank-ordered without any ties. In practice, ties commonly occur, and statisticians have devised ways to compensate for them. If there are relatively few ties simple adjustments to rank-order statistics usually suffice, but when there many ties an alternative to rank-order procedures may be needed. Such alternatives are commonly used for "partially ordered" scales, which arrange cases into a hierarchy of categories but make no distinctions within categories. With this sort of variable, every case is tied with at least one other case. Variables measured at the INTERVAL LEVEL identify differences between cases in such a way that they are not only ordered from lowest to highest but the SIZE of their differences can be stated in terms of a UNIT OF MEASUREMENT. Examples: Age measured in years and Income measured in dollars. DICHOTOMIES: If a variable divides cases into only two categories, it is often legitimate to assign arbitrary scores to its categories, e.g., 0 & 1, and treat it as an Interval Variable even if it was initially defined as a Nominal or Ordinal Variable. *SIZE NUMBER OF CASES is the number of distinct "units" that make up the sample, e.g., people or animals. NB: If more than one observation is made on each case, as in a before-after experiment, "Number of Cases" is not the same as "Number of Observations." If you do not know the exact N of Cases, use a best-guess estimate. Allowed range is 3 to 9999. If sample size is 10,000 or over, enter "9999." [WATSTAT uses N of Cases in deciding on procedures to recommend and in deciding when to warn you about potential limitations.] *N-OF-COMP-VARS What you enter here depends on the Analytical Focus you specified in Box 3. --------------------------------------------------------------------- If you chose "Univariate Analysis" as your Analytical Focus in Box 3, simply enter '1' here. If you chose "Association Between Variables" in Box 3, enter the number of INDEPENDENT VARIABLES you wish to use in your analysis. The ALLOWED RANGE of values you can enter is 1 thru 20. Enter '1' for a Bivariate Analysis and '2' or more for a Multivariate analysis. Use the maximum '20' if you have over 20 Independent variables. If you chose "Sub-Sample Differences" as your Analytical Focus in Box 3, enter the number of COMPARISON VARIABLES, as explained below. "COMPARISON VARIABLE" is WATSTAT's name for a variable whose function is to identify Sub-Samples to be compared. It is really an Independent variable, and may have any level of measurement. BUT due to its special function it is always treated as a set of Nominal categories. E.G., 'sex' would be a Comparison Variable if you compared the average memory-test scores of men and women. If you also divided the sample into age groups, in order to assess age-by-sex differences, 'age' would be a second Comparison Variable. The ALLOWED RANGE of values you may enter is 1 thru 20. At least one Comparison Variable is implied if you chose "Sub-sample Differences " in Box 3. Use the maximum '20' if you have over 20, but note that such a large No. of Comparison Variables may be unrealistic. [DON'T CONFUSE the No. of Comparison Variables with "No. of Categories" of those variables.] *SUBSAMPLE What you enter here depends on the Analytical Focus you specified in Box 3. --------------------------------------------------------------------- If you chose "Sub-Sample Differences" as your Analytical Focus in Box 3, AND you specified the No. of Comparison Variables in this Box, you must now tell WATSTAT how many Sub-Samples you wish to compare. The ALLOWED RANGE of values you can enter is 2 thru 99, BUT you must have at least 2 Sub-samples for each Comparison Variable you counted previously. Enter the maximum '99' if you have 100 or more Sub-samples. Be sure to count ALL the sub-samples. E.G., if comparisons are to be made by sex (2 categories) and age (5 categories), you'd enter '10' as the No. of Sub-Samples. Likewise, for experiments with multiple treatment factors, count EACH COMBINATION of treatments & treatment levels as a separate sub-sample. NB: For this item, "Sub-samples" are GROUPS of CASES OR OBSERVATIONS, so you should also count Before-After observations and other "repeated measures" as sub-samples, even though the same cases are involved.] If you chose "Univariate" or "Association Between Variables" in Box 3, AND IF you chose "Nominal," "Partially Ordered," or "Mixed: Nominal & Ordinal" levels of measurement Box 6, enter the TOTAL NUMBER OF CATEGORIES for all Nominal and "Partially Ordered" INDEPENDENT variables COMBINED. Don't count categories for Dependent variables nor for Independents that are "True" Ordinal or Interval variables, even if the latter are grouped. The ALLOWED RANGE of values you can enter is 0 thru 99. [Enter '0' if all Independent variables are measured at the Interval and/or 'True' Ordinal levels. Use '99' if you have 100 or more Categories.] Be sure to count ALL Categories; e.g., if data will be cross-tabulated by sex (2 categories) and age (5 categories), enter '10' as the No. of Categories. *MENU HELP [This message has 9 lines: keep scrolling down to read all of it] You can Select from the Menu in either of two ways: 1) use the arrow keys to highlight a choice and then press [ENTER], or 2) simply type the Number of your desired choice. Watch the Message Line at the bottom of the screen for usage information. Whenever it says "[F1] for Help" you can Press [F1] to bring up a Help Window like this. Always scroll down with the [ ] key until no new lines appear in a Help Window. [More ] When a Help Window is displayed, press [ALT] and [F1] together to expand it to full-screen size: this spares you from scrolling through longer help messages. To shrink a full-screen Help Window, press [ALT] & [F1] again. *RUNIT Choose this item if you're ready to start running WATSTAT. The first of a series of 'Choice Boxes' will pop onto the screen. Press [F1] now if you need help in using the Menu, [ESC] to return to it. *TUTOR Choose this option if you're new to WATSTAT. It explains how WATSTAT works and how to run it. It also explains how WATSTAT results should be used. Press [F1] now if you need help in using the Menu, [ESC] to return to it. *SOUND WATSTAT makes a 'trill' sound to alert you when a new Choice Box is popped onto the screen. Choose SOUND from the menu if you wish to turn the trill sound ON or OFF. [WATSTAT also beeps if you make an error, but this sound this sound can't be turned off.] Press [F1] now if you need help in using the Menu; press [ESC] to return to the Menu. *QUITIT Choose QUIT to terminate WATSTAT and exit to DOS. A window will pop onto the screen and ask for confirmation. All prior work is erased on exit. Press [F1] now if you need help in using the Menu, [ESC] to return to it. *COPYRIGHT COPYRIGHT 1991 BY HAWKEYE SOFTWORKS, 300 GOLFVIEW AVE., IOWA CITY, IA, 52246