143
OOM Software Manual
Observation Oriented Modeling
Software Manual
James W. Grice, Ph.D.
Oklahoma State University
Version 2
Software Release Date: November 12th, 2013
Updated: August 13th, 2015
Copyright © 2015
Contents
Introduction………………………………………… 3
Define Ordered Observations……………………… 6
Ordered Observations List Options……… 11
Auto Generate Options………………….. 13
Instructions/Distribution………………… 16
Build / Test Model…………………………………. 18
Overview and Initial Example………….. 18
Build Models……………………………. 22
Options………………………………….. 23
Randomization Test…………………….. 32
Output…………………………………... 34
Frequency or Proportional Models……... 41
Pairwise Rotation…………………………….……. 45
Options…………………………………. 46
Matching Analysis………………………...... 51
Options………………………………….. 52
Descriptive Statistics………………………………. 56
Pattern Analysis: Crossed Orderings…………… 58
Options………………………………….. 65
Randomization Test…………………….. 66
Output…………………………………... 67
Pattern Analysis: Concatenated Orderings…….. 69
Options………………………………….. 79
Randomization Test…………………….. 84
Output…………………………………... 85
Ordinal Analysis: Crossed Orderings……………… 86
Options………………………………….. 89
Randomization Test……………………. 91
Output…………………………………... 93
Ordinal Analysis: Concatenated Orderings………… 98
Options………………………………….. 108
Randomization Test…………………….. 113
Output…………………………………... 115
Efficient Cause Analysis…………………………… 118
Options………………………………….. 127
Randomization Test…………………….. 133
Output…………………………………... 135
Logical Ordered Observations…………………….. 136
Operators……………………………….. 140
Combine Units of Observations…………………… 142
Create Combination Orderings……………………. 145
Ordering/Case Combinations………….. 145
Group Combinations …………………... 147
1 Introduction
The purpose of this manual is to provide a brief overview of the different features and analysis routines in the Observation Oriented Modeling (OOM) software. In this regard it is meant to introduce the reader to various options in the software and to explain the output generated by these options. It also explains in plain language the logic of different analyses and the computations involved in generating different output. This manual is not meant to serve as a guide for building and testing integrated models nor is it meant to offer a complete guide for interpreting results generated by the different analyses. Still, careful study of this guide, along with viewing the instructional videos at http://www.idiogrid.com/OOM, should give the user a high level of comfort and confidence when using the OOM software.
The reader is encouraged to work through the examples included in this manual and in the videos. The data sets are included in the installation of the OOM software. Moreover, the reader is encouraged to experiment with his or her own data or with data constructed to have certain properties. Working with non-genuine or simulated observations is a good way to test the reader’s understanding of the software as well as the software’s capabilities. For example, the reader could generate pairs of ordered observations with a non-linear pattern of relationship and examine how the binary Procrustes rotation recovers the relationship.
The OOM software is constructed in a standard Windows format with a parent window and three child windows nested within: the Data Edit, Text Output, and
Graphics Output windows. These windows are layered in Figure 1.1, and a Main Menu can also be seen across the top of the parent window. The Data Edit window is currently active, or visible, in Figure 1.1.
Figure 1.1 OOM Parent and Child Windows
As a quick run through the program and an analysis, consider the following observations regarding smoking and lung cancer:
smoking cancer
person 1 No No
person 2 No No
person 3 No No
person 4 No No
person 5 No Yes
person 6 Yes No
person 7 Yes No
person 8 Yes Yes
person 9 Yes Yes
person 10 Yes Yes
File: SmokingCancerExample.oom
In OOM all observations must be represented with a number that can be entered into the Data Edit window. Clearly, observing whether or not a person smokes cigarettes or has developed lung cancer does not require the conceptualization of continuously structured quantitative qualities. The reliance on numbers for all ordered observations should not therefore be interpreted as assuming continuous quantitative structure in OOM; but instead, should be viewed as a clerical necessity in the software. For this example, 0 is used to represent “no” and 1 is used to represent “yes.” The observations as entered in the Data Edit window are shown in Figure 1.2. As can be seen, the ten persons form the rows of the observation matrix, and the two orderings form the columns. Zeros and ones are entered into the matrix to represent the observations.
The units of observations must next be defined. This is done in the Define Ordered Observations window which can be opened by selecting Edit: Define Ordered Observations from
Figure 1.2 Data Edit Window (snipped)
Figure 1.3 Define Ordered Observations Window
the Main Menu or by selecting the corresponding button from the toolbar (see Figure 1.1). Pausing the mouse over the buttons on the toolbar will briefly display their labels. Figure 1.3 shows the window with the smoking units of observation defined as:
{0} No
{1} Yes
The Cancer ordered observations are defined in the same way, and it should be pointed out that defining the units of observations correctly is critical in OOM. An entire chapter (Chapter 2) is therefore devoted to the Define Ordered Observations window.
Now that units of observation have been defined, analyses may be conducted. The standard analysis window in OOM is the Build / Test Model window listed under the Analysis Main Menu option. Figure 1.4 shows the window with the following expression being tested,
Smoking à Cancer.
Selecting the [OK] button to run the analysis sends the text portion of the results to the Text Output window and the graphics portion of the analysis to the Graphics Output window. Figure 1.5 shows the multigram generated from the analysis as it appears in the Graphics Output window.
Figure 1.4 Build / Test Model Window
Figure 1.5 Graphics Output Window with Multigram
2 Define Ordered Observations
Perhaps the most important window in OOM is the Define Ordered Observations window shown in Figure 2.1. It is in this window the user defines the units of observation that are the basis for the deep structures utilized by most of OOM’s procedures. Unlike other statistical programs, defining and labeling the units of observation is not simply a matter of convenience; rather, it is a necessity.
It can be seen that the window is separated into two sub-windows: the Ordered Observations list and Unit Definitions. The list of ordered observations is used to name the different orderings, define the numeric value that indicates missing observations, and set the decimal precision for which values are displayed in the Data Edit window. Several other options (viz., Min, Max, and Units) may be used in the process of defining units of observation. The unit definition sub-window is where the units of observation are actually defined and labeled, and several options are available to simplify this process. Because of the importance of defining the units of observation in OOM, the unit definitions sub-window also contains an edit box that reports simple instructions on how to define the units. These instructions can also be toggled to show a distribution (i.e., frequency histogram) of the observations as they are being defined. The distribution is important for insuring that all of the observations have been properly defined.
As with most of the chapters in this technical manual, the most expedient route for explaining the Define Ordered Observation window is via example. Consequently, we will consider 10 observations ordered according to 2 units of
Figure 2.1 Define Ordered Observations Window
Gender (Male/Female), 3 units of a subjective Rating of the U. S. President’s foreign policy (Disapprove, Neither Approve nor Disapprove, Approve), and 11 units of body (Temperature (98.0 to 99.0 with a single decimal of precision):
Gender Rating Temp
person_1 Male Approve 98.9
person_2 Male Disapprove 98.6
person_3 Male Disapprove 98.9
person_4 Male Disapprove 98.1
person_5 Male Neither 98.4
person_6 Female Approve 98.7
person_7 Female Approve 98.5
person_8 Female Neither 98.6
person_9 Female Approve 98.6
person_10 Female Approve 98.9
Clearly, the Gender observations are made through a discrete judgment of determining if a person is male or female. Nonetheless, in OOM numbers must be used to represent the units of observation. Given the nature of Gender, the choice of numbers to represent the observations is completely arbitrary; for example, 0 could just as easily be used as 100 or 70 to indicate a male. For the present purposes, 1 will be used to indicate a male and 2 will be used to indicate a female. The Rating observations are similarly discrete countable units and can be indicated by any numbers. Here, -1, 0, and 1 will be used to indicate the Disapprove, Neither, and Approve observations, respectively. The negative to positive values will serve a nice reminder of the apparent valence of the rating judgments (negative to positive). Lastly, body temperature is known as a continuous quantity and the values shown above can be entered “as is” in the OOM software and defined accordingly. The observations as they are entered into the Data Edit window thus appear as follows:
Gender Rating Temp
person_1 1 1 98.9
person_2 1 -1 98.6
person_3 1 -1 98.9
person_4 1 -1 98.1
person_5 1 0 98.4
person_6 2 1 98.7
person_7 2 1 98.5
person_8 2 0 98.6
person_9 2 1 98.6
person_10 2 1 98.9
File: DefineObservationsExample.oom
Turning now to the Define Ordered Observations window, Figure 2.2 shows the window as it will appear when all of the units of observation have been defined, when the Gender ordering has been selected, and the distribution has been toggled on.
Figure 2.2 Define Ordered Observations Window, Gender Defined
It can be seen that the following text appears in the Unit Definitions edit window:
{1} Male
{2} Female
This text defines the units of observations, NOT the Min and Max values in the observation list. The Min and Max values are only used in the Auto Generate options described below. It is the text in the Unit Definitions window that defines the observations upon which deep structure data matrices are constructed in OOM. The text “{1} Male” shows that the number 1 will be used to indicate a male. The brackets therefore enclose the number or numbers used to indicate a particular unit of observation, and the label appears to the right of the brackets. Similarly, the text “{2} Female” shows that the number 2 is defined and labeled as the indicator for a female.
The frequency distribution on the right side of Figure 2.2 shows that all 10 observations have been successfully defined as males or females. There are 5 males and 5 females in the data set. The text,
Male : [ 5]*****
Female : [ 5]*****
Total Number of Observations : 10
Number of Missing Observations : 0
Number of Units : 2
Observations to Categorize : 10
Categorized Observations : 10
Uncategorized Observations : 0
informs the user that all 10 of the observations have been accounted for in the definitions. If, for instance, the user were to mistakenly type the following text as the unit definitions,
{1} Male
{3} Female
then the following would appear in the distribution window:
Male : [ 5]*****
Female : [ 0]
Total Number of Observations : 10
Number of Missing Observations : 0
Number of Units : 2
Observations to Categorize : 10
Categorized Observations : 5
Uncategorized Observations : 5
Clearly, the 5 females in the data set have not been accounted for in the definitions. Their values were entered as 2’s and here miss-defined as 3’s. As mentioned above, it is in this way the distribution (frequency histogram) plays an important role in insuring the user has defined all of the units of observation properly.
Figure 2.3 shows a close-up of the ordered observations list in Figure 2.2. The Label for each ordering is determined and entered by the user and can be of any width and can include any characters. If the list of ordered observations is lengthy, they can be entered or changed in an edit box by selecting the [Edit Labels] button below the list (see Figure 2.2). Lists of labels can also quickly be copied from other programs (e.g., word processing, spreadsheet, or statistics programs) using the [Edit Labels] option.
Figure 2.3
Ordered
Observations
List
It can also be seen in Figure 2.3 that the Min, Max, and Units values for Gender are 1, 2, and 1, respectively. These values have no direct bearing on the actual unit definitions of the Gender ordered observations. They can be used, however, as aids in the automatic generation of units of observations.
Specifically, these values can first be set and then the [Single] button in the Auto Generate section of Figure 2.2 can be selected to automatically generate the units of observation with number labels based on the Prefix setting (in this case, “Unit=”). Doing so for Gender would yield the following default, automatically generated units of observation:
{1} Unit=1
{2} Unit=2
The [Single] automatic routine begins by creating a unit of observation from the Min value and labels it with the value affixed to the Prefix, in this case “Unit=1.” The routine then increments by one unit as defined in Units, in this case 1 and generates a second unit of observation with the label “Unit=2.” This process is incrementally repeated until the Max value is reached in the unit-generation process. For Gender, the process begins with 1 and increments by 1 to 2 at which point it stops, generating the text shown above.
While this process is convenient for generating units for orderings with numerous units of observation, in this example the number of units is only 2; moreover, the labels “Unit=1” and “Unit=2” are clearly not informative, so they can easily be edited to read “Male” and “Female” as originally shown above.
To reiterate, the purpose of the Min, Max, and Units values in the ordered observations list is to assist with the automatic generation of the units of observation in the Unit Definitions sub-window. The text in the Unit Definitions sub-window overrides these values. In other words, given the text in Figure 2.2; namely,
{1} Male
{2} Female,
the Min, Max, and Units values have no direct relevance to the ordered observations.