From Theory Construction to Testing


Like moths to lights, humans are attracted to organizing principles that give order, predictability, and meaning to life's random events. Through schemas, models, metaphors and myths, humans have sought to make sense of their world. Scientific theory is another of these ways by which experiences are organized and given coherence. In the abstract, science is nothing more than the exercise of developed perception. Its metaphors and theories direct our attention to particular phenomena and inform us about those things we are to ignore.

The first step of any science when approaching and conceptualizing any phenomenon is to think about it in terms of known metaphors. To describe the novel, one is forced to use known terms.

Consider the concept of society. What is it? How do the activities of thousands or millions of people come to be coordinated into some whole? At the turn of the century in the wake of Darwin's ideas, some social theorists conceptualized societies to be various organisms competing for survival on the inhabitable portions of earth.

Differing in their adaptability to changing "environments" (meaning not only volcanic eruptions, plagues and droughts but also new technological innovations, contacts with new cultures, and new religious or political ideologies), these social organisms are governed by the same survival-of-the-fittest laws as affect prairie dogs and Andean condors.

Applying this metaphor to the internal workings of a given society can give us some fairly nauseating insights. As a pig passes through a python so baby-boomers pass through American society, stretching the system (i.e., larger classrooms when school-aged, housing shortages when coming of age, fat layers of middle-management when middle-aged, etc.). As tissue grows around blood supplies so social organization grows around the flow of social resources, whether they be commerce, information, or knowledge. And with evolution--in other words, with the adaptations accruing in the struggles for survival--these parts and processes become increasingly complex, with greater differentiation and specialization. Hence, these theorists thought, where challenges for change are the greatest (i.e., those in the tropics didn't have to invent snowblowers and thermal underwear) you will find simple tribes evolving into complex nation-states. The "body politic" comes to have an elaborate communications system as its central nervous system, a political system for a brain (there will admittedly be controversy here), a vast military system for claws and fangs, and so forth. With the model in place, one can go on to perhaps liken crime to cancer (where portions of the social body feeds upon itself) or to parasitism (the animal kingdom is filled with non-productive thieves) and liken the growing bureaucracies as hardening of the arteries, slowing down decision-makings and abilities to respond to external threats.

Shifting from society to the individual, what metaphors come to mind when thinking about their feelings, thoughts and, most importantly, their behaviors? This choice of metaphor, in fact, remains the subject of considerable debate in social psychology. Can students' studying be likened to Pavlov's drooling dogs, with the first successful completion of homework being rewarded by candy and hugs and ultimately becoming motivated by a letter of the alphabet? In other words, is it all a matter of reinforcements? Can the way people make decisions be equated with the workings of a computer? Or might we think of all social action as being theatrical productions, where society is but a series of dramas ("The Family," "Work," etc.) wherein we act out our various roles ("Starring Me as the Oldest Brother," "And Also Featuring Jill as the Part-Time Receptionist," etc.).

In sum, at a minimum, metaphoric thinking provides order to what otherwise may be seen as random, unrelated phenomena. At a maximum, it provides an operating model or paradigm (from the Greek paradeigma, meaning an example, a model or a pattern) for the phenomena to be studied.

In the exercises that follow you will be investigating the relationships between selves and societies, how selves reside within society and how societies reside within selves.


What exactly does it mean when people say that poverty causes crime, that education diminishes racism, or that religion shapes individuals' premarital sexual activities? The matter isn't always as clear-cut as saying that when a rolling red billiard ball hits a black one that the red ball causes the black ball to move. But the inference is that when Y is in the presence of X that when X changes so will Y: when lungs are in the presence of cigarette smoke they are more likely to develop cancer than when not in the presence; when people are strongly religious they are more likely to say that premarital sex is wrong than those who are not religious.

Establishing that something truly causes something else is harder than it may seem. If two phenomena occur together it does not always mean that one is causing the other. The old example is the relationship between the number of fire trucks on a scene and the amount of damage that occurs--both increase together because of the workings of a third factor, here the size of the fire. The same holds true about the relationship between children's test scores and their shoe sizes--both tend to increase with age.

The world is filled with such relationships. In fact, whether or not a relationship occurs because of a causal connection or because of some spurious relationship can be the source of considerable controversy. Does televison viewing cause aggressive behavior and weaken moral norms? Or is the relationship simply an artifact of such variables as education, which we know determines the amount of television people watch and what they value? Are the racial differences observed in intelligence tests due to genetic differences or do they occur because of test-biases and socio-economic differences? Is the strong positive relationship between the increases in working mothers and in adolescent crime and violence causal, spurious, or independent phenomena?

The best way to test whether something causes something else is limited to the classical experimental design. Does, for instance, CAVITY-BLASTER really reduce tooth decay? To find out, you may choose to first measure the instance of cavities in two identical groups and then have the experimental group apply CAVITY-BLASTER while the control group continues brushing with their old toothpastes. Ideally, the mean number of cavities for individuals in both groups was originally the same. A few months later dentist Bob reinventories the cavity count and any difference in the mean number of cavities in the two groups can be attributed to the new paste.

When working with social surveys we normally cannot make such before-and-after measurements. But we can make inferences. For example, does having been raised by a single parent eventually lead one to remain single or postpone one's own marriage when an adult? This question seems to be most relevant nowadays as during your lifetime, for the first time in history, a married couple is as likely to be parted by divorce as by death. Further, according to the social prognosticators, more than one-half of American children are likely to experience the dissolution of their parents' marriage by the time they are 18.

If we had a survey where people were asked by whom they were raised and what their own marital experiences have been, our control group becomes those who lived with both parents while the experimental group is comprised of those raised by one. If we can make everything else "equal" between these groups--i.e., take into account their members' age (20-year-olds don't have save time to get married as 40-year-olds), sex (males continue to get married at older ages than females), and social class (children raised by single mothers more likely to have been raised in poverty than those raised by two parents, and individuals' social class of origins is highly predictive of their evental social class in midlife)--then whatever differences in singlehood or age of marriage can be inferred to be due to having been raised by one parent or two.

Thinking of the World in Terms of Variables

In the data sets that you will be working with there are numerous pieces of information or variables about each of the surveyed respondents. For instance, you will know their their sex, age, and marital status, the region where they live, how often they pray, and their attitudes toward such far-ranging issues as abortion, homosexuality, and the causes of poverty. Each of these variables will referred to by a distinctive name, such as SEX or PRAY.

Each variable sorts individuals into two or more mutually exclusive and totally inclusive categories. For example, the variable EDUCATION sorts people into four categories of educational background: those having 11 years or less of schooling, a high school diploma, some post-secondary education, and those with 4 or more years of college. Each individual falls into only one of these categories (because they are mutually exclusive) and there's a category to cover every person (the categories are totally inclusive--which is why there invariably is the "everything else" category OTHER).

With your CHIP software you will be looking at the ways in which these variables are related to each other. Researchers typically begin with a dependent variable, the phenomenon--e.g., happiness, divorce history, or attitude toward abortion--that they ultimately wish to explain. The dependent variable is, in turn, hypothesized to be caused by one or more independent variables.

Happiness, for instance, may be determined by such variables as marital status (e.g., married persons are happier than those divorced or widowed), group memberships (we may hypthesize that joiners are happier than loners), and income.

Often the relationship between one independent variable with a dependent variable is the central focus of researchers. For instance, the focus may be on the marital status --happiness connection.

From here the focus is on how that relationship is affected under varying conditions of additional independent variables. One may inquire how the marital status-happiness relationship differs for men and women, how it changes with age, or whether it is intensified the more religious or educated one is.

 Thinking in Terms of Causal Models of Variable Relationships

To represent what is causing what, to symbolically show the relationship between our dependent and independent variables, we will be be using causal models. These models are, in fact, diagrams of one's theory. These theories are not the product of armchair reflections with a glass of wine but rather built upon the ideas and findings of others.

Let's return to the question of whether having experienced as a youth the divorce of one's parents affects whether or not one eventually marries and, if so, whether one on average marries earlier or later than those coming from two-parent families. Suppose you discover when reviewing what other researchers have found that:

To put these findings into variable terms, we can summarize them thusly: The convention here will be to draw lines between related variables, going from the independent variable to the dependent variable with an arrow head point pointing to the dependent variable. Each variable has a "high" or + and and "low" or - category. For instance, high AGE may be those 60 and older while low age may be 25 and younger. The + category of PARDIVORCED may be yes (they divorced) and the - category means no, one's parents remained together. We will further note positive relationships--e.g., those where an increase from low to high of the independent variable produces an increase from low to high of the dependent variable--with solid lines and negative relationships with a dashed line. In the figure above, observe that two of the relationships are negative: the older one is the less likely one's parents were divorced; those from divorced families (the + category of PARDIVORCED had lower family incomes. The bottom relationship is positive: the higher the income of one's family of origin the higher the age at which one marries.

The figure on the right puts these variables together. Notice that a few additional lines have been added. Despite the recent trend toward later first marriages, it still remains the case that older generations married later than more recent ones, hence the solid (positive) line. And because of intergenerational upward mobility, the younger the individual the more likely he or she comes from better-off families.

Notice how the PARDIVORCED-AGEWED relationship now appears, embedded as it is within a complex system of variable interrelationships. There is, for instance, the possibility that the relation is in fact spurious because of AGE: the reason we may find children of divorce marrying earlier is because younger people are more likely to have divorced parents and are more likely to have married earlier to begin with. Those independent variables, like AGE, which are causally prior to the independent variable in question, namely PARDIVORCED, are known as antecedent variables. In addition, there are intervening variables, such as FAMILY$, which detour part of the influence of our independent variable PARDIVORCED on AGEWED, the dependent variable into which all causal arrows lead.

Speaking of metaphoric thinking, one way to conceptualize what's going on in such diagrams is to think of the lines as water pipes wherein relationships flow from antecendent variables and through intervening variables affecting the original pipeline (here PARDIVORCED-AGEWED) in question. The larger the relationship, the wider the pipe. If, for instance, the relationships between PARDIVORCED-FAMILY$ and FAMILY$-AGEWED are considerably larger than the PARDIVORCED-AGEWED connection, then the flow through the latter relationship "dries up" when taking into account FAMILY$. "So," you ask, "what is the proverbial bottom line? What have I gotten myself into?" Within a few chapters you will be able to assign values (or pipe sizes) to each of these lines taking into account the other variables in the system. If, for instance, the PARDIVORCED-AGEWED connection is actually a spurious relationship of AGE or if it is an artifact of the PARDIVORCED-FAMILY$-AGEWED conduit, then its line will be removed. In other words, in both cases the relationship has been explained away. If not, then you will be able to say precisely how large the direct influence is.


To illustrate the kinds of analyses you will be doing while simultaneously learning how to use the CHIP software, let's walk through the research problem developed above: What are the impacts of parental divorce on the marital histories of their children? Given the historical increases in divorce rates, what can we expect given what we know about trends when, if ever, people first go to the altar? Don't worry about understanding perfectly everything that we're about to do as we will more fully develop the various steps in the sections to come.

Before doing anything one must first get CHIP running and then open up a data set to get one's variables.


Getting Started

Turn on your system. You should see a prompt on the screen at the left, probably C:\>. That means there is a hard disk in your machine. Put your CHIP diskette into the drive (which may either by A or B). Assuming it is in drive A, type A:. Your prompt should now look like A:\>. Now you're ready.

To start the program type CHIP if you are using the Student Chip software or CH if you are using its related program, CHIPendale. The screen may go blank and a little time may pass but eventually you will see across the top of the screen the following boxes:


Each box contains a menu of things for the computer to do. These boxes can be accessed with the arrow controlled by a mouse or by pressing one of the F-keys located either on the top or left-side of your key board.

The first order of business is to get the data set for analysis. Each data set is a separate file that is also on your diskette. For those with a mouse, put the pointer over FILE and, while pushing down on its left key, drag the arrow to over the Open... O option and release the key. Otherwise, simply press the F1 key and with the down arrow key move until the Open command is highlighted and then press <ENTER>.

The O to the right of the Open command means that this process can also be accomplished by pressing the <Alt> and O keys simultaneously. There are several commands that have this short-cut and those that do have a letter following the command. Where this is the case, simply press <Alt> and the associated letter.

In the middle of the screen will appear a list of files (whose names end with .CHP) and possible directories (noted by <DIR>. The directories contain groups of files for the various sections of this text and can be accessed by highlighting one and then pressing the <ENTER> key. The directory AGING, for instance, contains files on the experiences of and changes associated with growing older. For this trial run, however, we will select the only file appearing: INTRO-1.CHP. Highlight this (either with the mouse arrow or by moving the down arrow key. Note that if you press the I key you will immediately get to this file.) and press <ENTER> (or the equivalent key on your mouse).

The file should now have been accessed and across the top of the screen you will see:


N = 2164

For each exercise the text will give you more information about the origin of the data and the specific variables in the file in the following format:

File: INTRO-1

Source: NORC 1990 & 1991 GSS


2 2 2 2 3

SEX: male, female

AGE: 25-35, 36+

PARDIVOR Constructed from following two questions: Were you living with both your mother and father around the time you were 16? (both parents, dad and stepmother, mom and stepfather, dad only, mom only, ...) and , if not, What happened? (death, divorce/separation, armed services, ...). Categories: parents were TOGETHER or DIVORCED

FAM$@16 Thinking about the time when you were 16 years old, compared with American families in general then, would you say your family income was--far below average, below average, average, above average, or far above average? Categories: BELOW AV, AVERAGE+

AGEWED Constructed from following two questions: Are you currently married, widowed, divorced, separated, or have you never been married? How old were you when you first married? Coded: NEVER, <23, 23+

Here we have the results from two years of the NORC General Social Survey with information on the responses given by 2,164 individuals. There are five variables. Beneath each are the number of its categories: not surprisingly, there are two sexes along with two categories of age (25-35, 36 and older), and so on. This "Info" can be obtained by highlighting the Command box (which can be accomplished by pressing the F2 key) and selecting the Info option. Note that Info also has a letter following it which means that you could have also obtained this information by simultaneously pressing the <Alt> and I keys.

The order of the variables listed is not random. Their order, in fact, is causal. Sex leads to age as females are more likely to be older than males (If it were the other way around, then one's sex would change with age!) As we found in previous studies, the older one is the less likely one's parents were divorced or separated when one was 16. Parental divorce also, according to others' arguments, can affect where one ends up in the social hierarchy (social class). Finally, social class determines when one first marries. We suspect the higher one's class the older one is when marrying.


How good are your predictive abilities? What percentage of Americans 25 and older, who had both biological parents alive when growing up, do you expect to have come from divorced families? If you believe the screaming headlines of the mass media, the figure must certainly be more than one-quarter as supposedly one-half of all marriages now ending in divorce and given the projections of one-third of all children born in the 1980s ending up in stepfamilies before they are eighteen. But subtracting number of divorces from number of marriages obscures several phenomena, such as multiple countings of those who divorce and remarry a number of times.

With these points in mind, your prediction:

1. % Americans 25+ with both parents alive when 16 but who did not live with both

2. And what percentage of Americans twenty-five and older do you predict have never married--not the percent single (who may have never married but also have divorced or are widowed) but those who never entered into a legal marital union. Some fodder for your calculations: In 1957, eight out of ten Americans claimed that an unmarried woman was "sick," "neurotic," or "immoral." As of 1992, 61% of all American adults were wed.

% Americans 25+ who have never married


3. Individuals whose parents had divorced before they were 16 are (circle one)

MORE          SAME          LESS

likely to have never married than their counterparts who lived with both biological parents.


4. Final prediction: Individuals whose parents had divorced before they were 16 marry than their counterparts from two-parent homes (circle one)


Next let's take a look at our marginal percentages, the percentage of individuals falling into the different categories of each of our variables. Repeat what you did to get the variable information but this time go the the All Marginals option. What you should see is the information below.

Observe, for instance, that 56.9% of our sample is comprised of females, 29.5% are 25-to-35 years-of-age, 12.8% of individuals when 16 years old had divorced or separated parents, and that nearly half our sample was married before aged 23 and 14.2% never married.


42.9 57.1 100.0%


25-35 36+ TOTAL
29.5 70.5 100.0%


87.2 12.8 100.0%


31.7 68.3 100.0%


14.2 49.6 36.2 100.0%

You will find that this information scrolls quickly by. To see what information has passed, with your mouse place the cursor in the box on the right and simply push it upwards. Should you be mouseless, simultaneously press the CTRL and S keys, which will allow you to use the PgUp and PgDn keys.

How accurate were your first two predictions? Of Americans who had both parents alive during their childhoods, only 12.8 percent--or about one in eight--did not live with both when they were sixteen. And about 14 out of every 100 Americans twenty-five and older have never married.

Cross Tabulation: Frequencies & Percentages

Marginals can be interesting--and we will be considering such statistics as the percentage of Americans who have had ESP and deja vu experiences--but they are only the first piece of our model. In particular, we are interested in how the marginal percentages of our dependent variable change within categories of our independent variables. For instance, are men and women equally likely (12.8% to be exact) to have come from divorced families? Are they equally likely to have remained single?

Let's first see if younger adults are more likely to have come from divorced families. To find out, once again go to the COMMAND option and this time select the Cross Tab routine (or, alternatively, simultaneously press the <ALT> + X keys). This is the master command of CHIP. Up comes the list of variables in the file. Our convention will be to first select our independent variable and secondly the dependent variable. First highlight variable AGE (either by moving the pointer to it or by pushing the down arrow  key and then press the <ENTER> box or, with the mouse arrow, select OK . Do the same thing to select the variable PARDIVOR.

You will now notice a different set of commands on the top of the screen: Options, Standardize, and Help. Go to the left-most, OPTIONS, box and select Frequency. You will see the following table:


36+ 1365 161 1526.0
25-35 522 116 638.0
ALL 1887 277 N=2164.0

In this table are our 2164 individuals. The right-most column are the frequency marginals of age, with 1526 people 36 and older and 638 aged 25 to 35 years. If you divide these numbers by 2164 you will obtain the marginal percentages reported above: 70.5% (1526/2164) are 36 and older and 29.5% (638/2164) are in the younger category.

The bottom row are the frequency marginals of whether or not one's own parents were divorced or separated: 277 (or 277/2164 = 12.8%) of our respondents came from a divorced family. The figures within the table are the numbers of individuals within the different categories of variables AGE and PARDIVOR For instance, 116 people are aged 25 to 35 and had divorced parents. The majority of cases, 1365 to be precise, are those aged 36 and older from parents who were together when the respondents were sixteen.

However, it is not clear whether our younger subjects are more likely than those older than 35 years to have come from divorced families. Although more older persons (161 vs. 116) are in the DIVORCED column, there are also more of them (1526 v. 638). To compensate for the differences in numbers of these two age categories, we can standardize by percentaging each row of the table so that each totals 100 percent. Put away your calculators--CHIP does this for you.

Return to the OPTIONS command and this time first select the Percent option and then the Percent Across. You then will see the following:

Sub option:

% Across

36+ 89.4 10.6 1526.0
25-35 81.8 18.2 638.0
ALL 87.2 12.8 N=2164.0

Yes, the younger age-group is indeed more likely to have come from divorced families--7.6 percentage points (18.2 - 10.6) more likely to be precise. Another way of expressing this difference is to say that those 25 to 35 are over 70 percent more likely (18.2/10.6) to be from "broken" homes than those aged 36 and older.

If there was no relationship between age and whether or not one's parents were divorced, the percentages in the 36+ and 25-35 rows would be identical to the All row: 12.8% of both age groups would have had divorced parents. As you will see, the greater the differences in the category percents and the All or Total percents, the larger the variable relationship is--in our example, the older the respondent the less likely his or her parents were divorced.


Hey, I'm a Visual Person, Let's See the Relationship: Graphing

Another way of looking at our relationship between age and parental divorce is to go to the Plot command and select Slope.  

There first appears a menu box titled "Categories" with the options All, Together and Divorced.  These are the categories of our dependent variable PARDIVOR to be graphed. Select Divorced.  You will then be asked to "Enter Vertical Range," in other words, should the vertical axis of the graph be 0 to 100% or do you wish to "blow it up" (size-wise, that is). Since only 12.8% of our total sample had divorced parents, keep zero as the minimum value and with the down-arrow key (do not press the <ENTER> key here or the defaults will be kept) move down to the maximum value line and enter 20. Move the mouse arrow to OK and click (or simply press the <ENTER> key). You should now see a plot. Because of our magnification, the age difference looks fairly impressive. Indeed, those 36 and older are considerably less likely to have had divorced parents than those 25 to 35.

Percentage Differences

It's time to return to the original question at hand: Does parental divorce have any effect upon their children's marital history?  

To clear away the previous analyses select the Exit command in the Options box (or, alternatively, simultaneously press the <ALT> and E keys). CHIP is ready for another pair of variables to crosstabulate. Note that the top boxes have returned to File, Command, Modify and Help choices.

Once again select the Cross Tab command within the Command box and then first select PARDIVOR followed by AGEWED as your two variables. Again, the top boxes return to the Options, Plot, Standardize, and Help choices. Under the Options box this time select Percent Diff.  

As Cross Tab is the most frequently used Command in CHIP so Percent Diff will be your most frequently used Option. As you will discover, all advanced analyses will be based on this procedure. 

Once selected, up pops in the middle of the screen another menu. CHIP wants to know what categories of the independent variable, here PARDIVOR to substract from what. For Group 1, the positive categories, highlight DIVORCED and either press <ENTER> or, with the mouse, click on the OK box. Then select No More. You will then be asked for the negative categories in Group 2. Select TOGETHER. Since variable PARDIVOR only has two categories, CHIP will now do its thing. Below is what you will see:



Group 1 Positive categories

Category: DIVORCED

Group 2 Negative categories

Category: TOGETHER

DIVOR(+) 20.9 50.5 28.5 277.0
TOGET(-) 13.2 49.4 37.3 1887.0
ALL 14.2 49.6 36.2 N=2164.0
% Diff 7.69 1.10 -8.79
d/sigma d 3.00 .34 -3.00

Replacement index = 8.79

Notice that like the Percent Across option, each row adds up to 100%. Of the 277 subjects whose parents divorced, 20.9% of them never married, 50.5% married before they were 23 years old, and the remaining 28.5% married when they were 23 or older.

What's different here are the % Diff and d/sigma d rows and the Replacement index. The percent difference row is the result of subtracting the positive (+) category, which we selected as DIVORCED, with the negative (-) category, TOGETHER. The 7.69 means that those of divorced parents are 7.69 percentage points more likely to have never married than their counterparts from two-parent families. The 8.79 means that those whose parents were together when they, the children, were 16 years old are 8.79 percentage points more likely to have married later than those whose parents were divorced.

The d/sigma d numbers tell you how statistically significant the percentage differences above really are. Recall that we are looking at a sample of adult Americans, and we could literally draw trillions of samples with 2164 individuals from the 150 million-or-so Americans aged 25 and older. Not surprisingly, we could expect these percentage differences to fluctuate. What the d/sigma d numbers tell you is whether the percentage difference would hold up. As a rule of thumb, any value of 1.96 or larger means that the likelihood of the percentage difference actually being zero is less than 5 in 100. If this is the case, as it is for two of our three differences, the values appear shaded.

Though we won't be talking much about it, the Replacement index simply means the percentage of cases that would have to be moved across the various cells to eliminate the relationship, to make each row of the table appear identical to the All or Total row.

So what does the data tell us? It says that children of divorced parents are significantly (this word has a precise statistical meaning) more likely to have never married than their two-parented counterparts who, in turn, are significantly more likely to have married later on.

Question: Is this relationship equally true for men or women? To find out, let's create two PARDIVOR-AGEWED tables: one made up only of women and the other only with men.


Return to the OPTIONS box at the upper left of your screen and select the Control C option (which can also be accomplished either by typing <ALT> + C or by pressing the F1 key and cursoring down to Control and pressing <ENTER>).

In the center of the screen appears a variable list. Notice that PARDIVOR and AGEWED are missing as they are the variables already selected for our table. Select SEX and click on the OK box (or press <ENTER>) and then select the No More option and again click on OK. You now have told CHIP to examine the PARDIVOR-AGEWED relation controlling for (in other words, looking at the relationship within categories of) SEX.

Once again go to the OPTIONS box and select Percentage Diff command and then the Subtables option. As we did before, select DIVORCED as the positive Group 1 category and TOGETHER as the negative group 2 category. You should see the following, first the table with only males and then, with a press of the <ENTER> key, the table with only females.



DIVOR(+) 25.8 38.7 35.5 124.0
TOGET(-1) 15.2 36.4 48.4 805.0
ALL 16.6 36.7 46.7 N=929.0
% Diff 10.65 2.31 -12.96
d/sigma d 2.58 .49 -2.79

Replacement index = 12.96



NEVER <23 23+ Total

DIVOR(+) 17.0 60.1 22.9 153.0
TOGET(-) 11.8 59.1 29.0 1082.0
ALL 12.5 59.3 28.3 1235.0
% Diff 5.16 .98 -6.14
d/sigma d 1.62 .23 -1.68

Replacement index = 6.14

Upon quick examination you note that there are shaded % Diff only for males, which means that only among men are that there statistically significant differences between the divorce history of their parents and the likelihood of these men never marrying or marrying late. Parental divorce appears to have less effect on the marital history of the fairer sex.

Modifications of Variables: Elimination of Variable Categories

We still haven't fully addressed the hypothesis about whether children of divorce marry earlier or later than their counterparts from "intact" families. Let's eliminate the NEVER married category of AGEWED so that we can consider just those folks who have married.

To do this, either select the Exit command within the Options box (or simultaneously press the <Alt> and E keys) to return to the original screen with the File, Command, Modify and Help choices. This time, select the Modify box and the Omit option. Up comes that variable menu. Select variable AGEWED. Next to appear are the categories of AGEWED. Select NEVER (the category we will be omitting) and then No More. That's it! Never have never-married individuals disappeared so quickly! In fact, they will be gone from all subsequent analyses until you reopen this file.

Once again it's crosstabulating time. Select the Cross Tab command and enter PARDIVOR and AGEWED. Select the Percent Diff and, again, make DIVORCED the positive group 1 category and TOGETHER the negative group 2 category. You will see:



Group 1 Positive categories

Category: DIVORCED

Group 2 Negative categories

Category: TOGETHER

<23 23+ TOTAL
DIVOR(+) 63.9 36.1 219.0
TOGET(-) 57.0 43.0 1637.0
ALL 57.8 42.2 N=1856.0
% DIFF 6.93 -6.93
d/sigma d 2.00 -2.00

Replacement index = 6.93

Compare this table with that above when the NEVER category was included. Instead of 50.5% of individuals of divorced parents marrying before they turned 23, the figure here is 63.9%.  The reason? By eliminating the never-marrieds our percentages here only apply to those who have married. But the conclusion remains the same: Those coming from divorced households are significantly more likely to marry earlier than those whose biological parents remained together.  

For a more complete analysis of this causal model click here.


By now, many tables have passed across your screen. Perhaps there are earlier tables that you wish to revisit. CHIP has several solutions.  

You may have noticed that two of the commands associated with the File box include Log and End Log. By selecting Log you have the ability to store all of your activities in a file that can later printed out, such as through your word processor. If selected you will see a small box asking you to name your output file. You can, for instance, enter B:\div-runs (or whatever 8-letter name instead of "div-runs" that you choose) and the file will be stored by that name on your B (or whatever drive you have besides the one where your CHIP diskette is located) drive. Do not store your output on your CHIP file, which in this example may be located in the A drive. To remember what data set is associated with your analyses, it is recommended that you open the log prior to selecting the Open option.

If you do not have a log open and if you have not selected the Clear option, which is like a giant eraser, you can still return to earlier tables. If you have a mouse, you can take advantage of the small box on the right of the screen between the up and down arrows. With the arrow on this box and pushing the left mouse key, this box can be moved up or down and when so doing you can scroll through your runs. There's no need to go to the end before making new commands as they will automatically be appended to the end of your output.

If you should be mouseless
, simultaneously press the CTRL and S keys. A small S will then appear in the upper right-hand corner of the screen. This allows you to use the up and down arrows on your keyboard or the PgUp and PgDn keys to move through your record.

Return to Methods Section of A Sociological Tour Through Cyberspace