The Parent/Family Involvement in Education (PFI) and Civic Involvement (CI) components of the NHES:96, which are the basis of this report, employed a sample of children and youth from age 3 through 12th grade. Up to three instruments were used to collect information included in this report. The first instrument was a set of household screening items (Screener) administered to an adult member of the household, which was used to determine whether any children of the appropriate ages lived in the household, to collect information on each household member, and to identify the appropriate parent/guardian to respond for the sampled child. For sampling purposes, children residing in the household were grouped into younger children, age 3 through grade 5, and older children, in grades 6 through 12. One younger child and one older child from each household could have been sampled for the NHES:96. If the household contained more than one younger child or more than one older child, one from each category was randomly sampled as an interview subject. For households with youth in 6th through 12th grade who were sampled for the survey, an interview was conducted with the parent/guardian most knowledgeable about the care and education of the youth, and following completion of that interview and receipt of parental permission, an interview also was conducted with the youth.
For the NHES:96, item nonresponse (the failure to complete some items in an otherwise completed interview) was very low. For some items in the interview, a response of don't know or refused was accepted as a legitimate response. Using an imputation method called a "hot-deck procedure" (Kalton and Kasprzyk, 1986), responses were imputed for missing values (i.e., "don't know" or "refused" for items not specifically designated to have those legitimate response categories, or "not ascertained"). As a result, no missing values remain. Item nonresponse rates for variables in this report are generally less than 2 percent. The following items used in this report had nonresponse rates greater than 2 percent. For each item, the variable name, a description of the variable, number of eligible respondents, and item nonresponse rate are shown.
In general, it is difficult to identify and estimate either the amount of nonsampling error or the bias caused by this error. In the NHES:96, efforts were made to prevent such errors from occurring and to compensate for them where possible. For instance, during the survey design phase, focus groups and cognitive laboratory interviews were conducted for the purpose of assessing respondent knowledge of the topics, comprehension of questions and terms, and the sensitivity of items. The design phase also entailed CATI instrument testing and an extensive, multi-cycle field test.
An important nonsampling error for a telephone survey is the failure to include persons who do not live in households with telephones. About 93 percent of all students in kindergarten through 12th grade live in households with telephones. Since the sample for the NHES:96 was drawn from households with telephones, the estimates were adjusted using control totals from the Census Bureau's Current Population Survey (CPS) so that the totals were consistent with the total number of civilian, noninstitutionalized persons in all (telephone and nontelephone) households.
Another potential source of nonsampling error is respondent bias. Respondent bias occurs when respondents systematically misreport (intentionally or unintentionally) information in a study. There are many different forms of respondent bias. One of the best known is social desirability bias, which occurs when respondents give what they believe is the socially desirable response. For example, surveys that ask about whether respondents voted in the most recent election typically obtain a higher estimate of the number of people who voted than do voting records. Although respondent bias may affect the accuracy of the results, in the voting case the estimate of the number who voted, it does not necessarily invalidate other results from a survey. If there are no systematic differences among specific groups under study in their tendency to give socially desirable responses, then comparisons of the different groups will accurately reflect differences among the groups. In this report, there may be a tendency for respondents to say that they participated in a school activity when they did not. There is no a priori reason, however, to believe that parents in two-parent families are more likely than those in single-parent families or that mothers are more likely than fathers to give the socially desirable response. Thus, it is likely that contrasts in this report reflect true differences between fathers and mothers and parents in single-parent and in two-parent families.
Another form of respondent bias occurs when respondents give unduly positive assessments about those close to them. For example, parents may give rosier assessments about their children's school experiences than might be obtained from school records or from the children themselves. It is possible that parents who are highly involved in their children's schools are more likely than those who are not so involved to say that their children are doing well in school or that their children enjoy school. However, it is also possible that parents who are highly involved in their children's schools have more information than those who are less involved on which to base their reports. This information could be positive or negative. Thus, it is equally conceivable that parents who are highly involved in their children's schools are less likely than other parents to give rosy assessments of their children's school experiences. Readers should be aware that respondent bias may be present in this survey as in any survey. It is not possible to state precisely how such bias may affect the results.
The standard error is a measure of the variability due to sampling when estimating a statistic. Standard errors for estimates presented in this report were computed using a jackknife replication method. Standard errors can be used as a measure of the precision expected from a particular sample. The probability that a complete census count would differ from the sample estimate by less than 1 standard error is about 68 percent. The chance that the difference would be less than 1.65 standard errors is about 90 percent, and that the difference would be less than 1.96 standard errors, about 95 percent.
Standard errors for all of the estimates in this report have been calculated and are available from NCES upon request. These standard errors can be used to produce confidence intervals. For example, it is estimated that 55 percent of fathers in two-parent families with children in kindergarten through 5th grade attended a meeting at their child's school, and this statistic has a standard error of 0.54. Therefore, the estimated 95 percent confidence interval for this statistic is approximately 54 to 56 percent.
All of the estimates in this report are based on weighting the observations using the probabilities of selection of the respondents and other adjustments to partially account for nonresponse and coverage bias. These weights were developed to make the estimates unbiased and consistent estimates of the national totals. In addition to properly weighting the responses, special procedures for estimating the statistical significance of the estimates were employed because the data were collected using a complex sample design. Complex sample designs, like that used in the NHES, result in data that violate some of the assumptions that are normally required to assess the statistical significance of the results. Frequently, the standard errors of the estimates from the survey are larger than would be expected if the sample was a simple random sample and the observations were independent and identically distributed random variables. WesVarPC was used in this analysis to calculate standard errors for both bivariate estimates and regression analyses.
Replication methods of variance estimation were used to reflect the actual sample design used in the NHES:96. A form of the jackknife replication method was used to compute approximately unbiased estimates of the standard errors of the estimates in the report. The jackknife methods were used to estimate the precision of the estimates of the reported national totals, percentages, and regression parameters. To test the differences between estimates, Student's t statistic was employed, using unbiased estimates of standard errors derived by the replication methods mentioned above.
As the number of comparisons at the same significance level increases, it becomes more likely that at least one of the estimated differences will be significant merely by chance, that is, it will be erroneously identified as different from zero. Even when there is no statistical difference between the means or percentages being compared, there is a 5 percent chance of getting a significant F or t value from sampling error alone. As the number of comparisons increases, the chance of making this type of error also increases. A Bonferroni adjustment procedure was used to correct significance tests for multiple comparisons. This method adjusts the significance level for the total number of comparisons made with a particular classification variable. All the differences cited in this report are significant at the 0.05 level of significance after a Bonferroni adjustment. For example, the total number of comparisons for the race/ethnicity variable is six (i.e., white, non-Hispanic vs. black, non-Hispanic; white, non-Hispanic vs. Hispanic; white, non-Hispanic vs. other race; black, non-Hispanic vs. Hispanic; black, non-Hispanic vs. other race; Hispanic vs. other race). Thus, the significance criteria for each race/ethnicity comparison is adjusted to p=0.0083 (i.e., .05 / 6).
Meeting=.Essentially, respondents who received the second version that consisted of two questions were said to have attended a general school meeting if they had responded yes to either one of the two types of meetings. They were said not to have attended a general school meeting if they had not attended either type of meeting.
If FSMEETNG=1 or (FSBAC=1 or FSATTPTA=1) then Meeting=1;
else if FSMEETNG=2 then Meeting=2;
else if (FSBAC=2 and FSATTPTA=2) then Meeting=2;
Number of school activities parents participated in. Information on whether any adult had attended each of the four types of school activities and which adult had attended was used to create an indicator of maternal involvement and an indicator of paternal involvement. For each activity that either the mother or both parents had attended, the indicator of maternal involvement (Cntmom2) was increased by one. Similarly, for each activity that the father or both parents had attended, the indicator of father involvement (Cntdad2) was increased by one. Cntmom2 and Cntdad2 range from 0 (no activities attended) to 4 (all four activities attended). Parallel variables were created for nonresident fathers and mothers who had had contact with their children in the past year.
High maternal and paternal involvement. The variables measuring high maternal and paternal involvement were based on Cntmom2 and Cntdad2. Two dichotomous variables were created that were assigned a value of 1 if the parents had attended three or four of the activities and were assigned a value of 0 if they had attended none, one, or only two of the activities. Parallel variables were created for nonresident fathers and mothers who had had contact with their children in the past year. For the nonresident parents, however, the dichotomy was between nonresident parents who had participated in two or more activities in their children's schools versus those who had participated in none or only one activity.
Children's contact with their nonresident parents. The measure on children's contact with their nonresident fathers and mothers has the following categories:
Nr1stat=.;
If NRLIVAR1=2 or NRLIVAR1=3 then Nr1stat=1;
Else if (NRLIVAR1=4 or NRCONTA1=3) then Nr1stat=4;
Else if (NRLIVAR1=5 or NRLIVEV1=2 or NRCONTA1=4) then Nr1stat=3;
Else if LASTCON1 gt 12 then Nr1stat=2
Else if 1 <= LASTCONT1 <= 12 then Nr1stat=1;
If NR1STAT=. then do;
If NRCONTA1=2 and NRLSTCO1=2 then Nr1stat=3;Children were said to have had contact with their nonresident parent within the last year if any of the following were true:
Else if NRCONTA1=2 and NRLIVEV1=1 then Nr1stat=2;
Else if NRLIVEV1=0 then Nr1stat=3;
End;
If the respondent reported that the nonresident parent was deceased, the child was considered not to have a nonresident parent.
A parallel variable, Nr2stat, was created for the second identified nonresident parent. Once these two variables were created, two additional variables (Momstat and Dadstat) were created that took the value of Nr1stat or Nr2stat depending upon which one referred to the nonresident mother or the nonresident father for a particular case.
Ever repeated a grade. This dichotomous variable is based on SEREPEAT. It takes a value of 1 if the child has ever repeated a grade and a value of 0 otherwise.
Enjoys school. This dichotomous variable is based on SEENJOY. It takes a value of 1 if the parent agrees or strongly agrees with the statement that "child enjoys school" and a value of 0 otherwise. The question was not asked of children in kindergarten, so the variable is set to missing for them.
Ever suspended or expelled. This dichotomous variable is based on SESUSEXP. It takes a value of 1 if the parent reports that the child has ever been suspended or expelled and a value of 0 otherwise. The question on suspension or expulsion was only asked about children in grades 6 through 12, so the variable is set to missing for all other children.
Participates in extracurricular activities. Parents of children in kindergarten through grade 5 were asked whether their children had participated in any school activities such as sports teams, band or chorus, or safety patrol. They were also asked whether during the school year the children had participated in any activities outside of school, such as music lessons, church or temple youth group, scouting, or organized team sports, like soccer. If the parent reported yes to either of these questions, the child was said to have participated in extracurricular activities, otherwise the child was said not to have participated. Children in grades 6 through 12 were asked the same two questions. If the 6th through 12th graders reported that they had participated in school or non-school activities during the school year, they were said to have participated in extracurricular activities, otherwise they were said not to have participated.
An example will help clarify the concepts. The odds that fathers in two-parent families are highly involved in the schools of their 6th through 8th graders and of their 9th through 12th graders can be calculated using the descriptive information presented in figure 3. According to figure 3, 25 percent of fathers in two-parent families are highly involved in their 6th through 8th graders' schools and 23 percent are highly involved in their 9th through 12th graders' schools. The odds that fathers are highly involved in their 6th through 8th graders' schools are calculated as follows: 0.25/(1-0.25)=0.33. Similarly, the odds that fathers are highly involved in their 9th through 12th graders' schools are 0.23/(1-0.23)=0.30. The odds ratio, 0.33/0.30, measures the change in the odds that fathers are highly involved in their children's schools that is due to the children's grade level. In this case, the odds that fathers are highly involved in their children's schools are 1.1 times as large for fathers of 6th through 8th graders as they are for fathers of 9th through 12th graders. This can also be expressed as a percent change in the odds calculated as (odds ratio-1)*100. A positive value indicates a percent increase in the odds and a negative value indicates a percent decrease in the odds. Thus, one can also say that the odds that fathers are highly involved in their children's schools are 10 percent greater for fathers of 6th through 8th graders than they are for fathers of 9th through 12th graders. This does not mean, however, that fathers of 6th through 8th graders are 1.1 times more likely or are 10 percent more likely to be highly involved in their children's schools than fathers of 9th through 12th graders\2\. In this example, the relative risk or relative probability that they are highly involved is 0.25/0.23 or 1.09, which can also be expressed as a percent change in the relative risk, as follows: [(relative risk -1)*100=9]. In this case, the odds ratio and the relative risk are close. This is not always the case, however. Odds ratios will always overstate the difference in relative risks. It is always true, however, that whenever odds ratios are greater than 1 so is the relative risk. Similarly, whenever odds ratios are less than 1, so is the relative risk.
The reason that odds ratios are frequently used to summarize the results of logistic regression models is because odds ratios are easy to obtain and do not depend upon the values of the other variables in the model. Probabilities, on the other hand, change depending upon where on the logistic regression curve they are evaluated (that is, they depend upon the values of the other variables in the model).
FOOTNOTES:
[1] Not quite 3 percent of children with nonresident parents actually lived most of the school year with that parent.
[2] In trying to understand the influence of specific factors on the likelihood that an event will occur, it is important to control for potentially confounding factors. According to the results in table 3, after controlling for the other factors in the model, the adjusted odds that fathers are highly involved in their children's schools are 28 percent lower, rather than being 10 percent higher, for fathers of children in grades 6 through 8 relative to those in grades 9 through 12. The change in the interpretation highlights why it is important to control for potentially confounding factors.
Summary and Discussion
References
