Chapter 3: Research Method
The problem addressed in this study
was a lack of immigrant and refugee parent involvement in Head Start and Early
Head Start programs to support their children’s education (Hindman et al.,
2012; Manz et al., 2014). Specifically, this study addressed the lack of knowledge
regarding what variables relate to high and/or low parent involvement and identification
of those that served as barriers to or promoters of parent involvement in this
population. Some of the barriers for immigrant and refugee families were likely
their pre-conceived beliefs about education (Cheatham & Ostrosky, 2013) as
well as differences due to language and culture (Cheatham & Ostrosky,
2013). The variables, as defined within the theory of planned behavior (TPB)
that contribute to parental intentions and were included in the study attitudes
and beliefs, subjective norms, and perceived behavioral control (Ajzen, 1991;
Bracke & Corts, 2012; Perry & Langley, 2013). These variables had not
been thoroughly examined in the context of immigrant and refugee parents such
that this knowledge could be applied to help devise parental interventions for
these parents (McGregor & Knoll, 2015) in order to improve parental
intention for involvement in their children’s formal schooling. Without further
understanding of how these variables relate to and predict parental intentions
for involvement, strategies employed by these programs to increase the parental
involvement of immigrant and refugee families may be less effective and the
families may not take full advantage of these programs (Lee & Zhou, 2014). Identifying
determinants of parent involvement, or lack thereof, in immigrant and refugee
populations could subsequently be used to develop interventions (Hindman et
al., 2012).
The purpose of this quantitative, descriptive
and correlational study was to assess key variables, posed in the TPB as
possible determinants of parent intention for school involvement behavior
(i.e., parental attitudes and beliefs, subjective norms, and perceived
behavioral control) and ascertain whether they were significantly related to
and could predict the reported intentions of involvement of immigrant and
refugee parents in their children’s early childhood education programs. By
assessing the constructs that were pivotal to TPB, an expansion of this theory
within the context of immigrant and refugee families with children enrolled in
Head Start or Early Head Start programs was fulfilled. Using a survey
instrument to collect data, the goals included the following: (1) identify
parental attitudes and beliefs, subjective norms, perceived behavioral control,
and parental intentions regarding parent involvement, and (2) investigate how,
if at all, these variables correlated with and served to predict parent
involvement, as TPB would suggest, in this population. The dependent variable
was parental intentions for involvement, and the predictors were the
determinants of behaviors as outlined in TPB (i.e., parent attitudes and
beliefs, subjective norms, and perceived behavioral control) as reported by
immigrant and refugee families with children enrolled in Head Start and Early
Head Start programs.
The target population
was 800-1000 parents/caregivers who were foreign born, i.e., first-generation
immigrants or refugees (Krogstad, 2015; Winsler et al.,
2014), were living in Southern California, and had children enrolled in either
an Early Head Start or Head Start program. A census of this population was
conducted with the goal of obtaining a sample of 122 parents/caregivers whose
children participated in these programs. A power analysis using G-Power
software yielded an estimated sample size of 110 for a linear regression with
three predictors (power of 0.8, type one error of 0.05, and medium size effect)
(Faul, Erdfelder, Buchner, & Lang, 2009). Parents were asked to complete
the Parental Involvement Project
(PIP) Parent Questionnaire (Appendix A). It was a 57-item survey that had been
found reliable and valid for measuring attitudes and beliefs (24 items),
subjective norms (6 items), perceived behavioral control (17 items), and
parental intentions (10 items). All items were evaluated using six-point Likert
scales. Data collection provided an opportunity to describe how these variables
presented in immigrant and refugee families and examined their ability to
predict parent intentions towards involvement using the three predictor
variables of attitudes and beliefs, subjective norms, and perceived behavioral
control. The findings serve to further inform the TPB about its applicability
to immigrant and refugee parents who had children attending one of the Head
Start or Early Head Start programs. Step-wise multiple regressions were used to
assess the significance of the contributions of each predictor to explain the
variation of the dependent variable (Field, 2013). Survey methods were employed,
as they were the most appropriate method for collecting quantitative data to
measure study variables in the TPB model to ascertain whether they were significantly
related to and ccould predict the involvement of immigrant and refugee parents
in their children’s early childhood education programs (Girardelli & Patel,
2016). This chapter includes discussion of the research design,
population/sample, materials/instrumentations, operational definations of variables,
study procedures, data collection and analysis, assumptions, limitations,
delimitations, ethical assurances, and ethical guarantees that apply in the
proposed study.
Research
Design
A quantitative, descriptive and
correlational design was employed to answer the research questions. A
quantitative design was the most appropriate method since it generated numerical
data that were analyzed to determine relationships between the multiple
predictor variables in TPB (Park & Park, 2016). A regression model was used
to investigate whether the TPB determinants of behavior were significantly
related to and predicted the parental intentions for involvement of immigrant
and refugee parents in their children’s early childhood education programs. The descriptive and correlational design allowed
for assessing the key variables in TPB such as the determinants of behavior
(i.e., parental attitudes and beliefs, subjective norms, perceived behavioral
control, and parental intentions) and determining whether they were
significantly related to and predicted the intentions of immigrant and refugee
parents for being involved in their children’s early childhood education programs.
The approach employed in this study was an appropriate technique for testing
whether TPB constructs fit a model for an observed set of correlations among
the constructs in the model (Girardelli & Patel, 2016).
A cross-sectional design was used
to collect data to assess variables in TPB gathered at a single point in time
using valid and reliable instruments (Trochim & Donnelly, 2008). A
cross-sectional survey required less dedication from research participants,
took less time to complete, and did not contain many obstacles related to
finding and maintaining a sample population (Leedy & Ormrod, 2010) as
opposed to a longitudinal study that must take place across at least two waves
of times (Trochim & Donnelly, 2008). This method of data collection was
equated with surveys that must be carefully designed prior to the research
occurring (Fowler, 2009). This design allowed collection of data that could
measure variables quantitatively for statistical analyses from a sample that could
be generalized to a target population, while remaining objective, separated
from the subject matter, unbiased, and value-free (Smith, 2015). The
correlational research design sought to control preferences, so those facts,
instances, and phenomena were understood in an objective way (Park & Park,
2016). The strength of this research design was in the information that could be
reported in the form of numbers and could test a formulated hypothesis prior to
the actual collection of data (McCusker & Gunaydin, 2014). McCusker and
Gunaydin pointed out that when using this design, the extraction of information
in a larger volume and emphasis on statistical information rather than
individual perceptions. The weakness of the survey method occurs when the answers
from sampled respondents are not accurately measurable and become erroneous. To
avoid this problem, a reliable and valid survey was used (Fowler, 2009). A
survey designed by Hoover-Dempsey, Walker, Jones, and Reed (2002) was employed.
The validity and reliability of the survey were previously established in
studies by Hoover-Dempsey, Walkers, and Sandler (2002; 2005). The survey
consisted of 57 questions based on a six-point Likert scale assessed the
following variables: attitudes/beliefs, subjective norms, perceived behavioral
control, and parental intentions. The survey takes approximately 30 minutes to
complete and was returned to the researcher in person or by mail to receive a
gift card. By using a survey design, there were some advantages such as
inexpensive unit costs (Fowler, 2009); however, some disadvantages included
that parents did not always submit the surveys via mail within the timeframe given
(Fowler, 2009). The following section describes how the sample was obtained in
this study.
Population and Sample
The population for this study
included all immigrant and refugee parents or caregivers from All Kids Academy
Head Start and Early Head Start programs (AKA-HS/EHS). According to AKA HS/EHS’s
2015 annual reported, a subgroup of immigrant and refugee parents or caregivers
with ethnic and racial diversity for potential generalizability was identified
(Hindman, Miller, Froyen, & Skibbe, 2012). In 2015, AKA Head Start programs
served 1,167 children from low-income families. According to the Poverty
Guidelines published by a federal government, they were eligible for HS and EHS
services. All AKA Head Start centers provided full day and part day child care
services to families year-round for 10.5 hours per day, Monday through Friday
from 7:00 am to 5:30 am, and 3.5 hours per day, Monday through Thursday from
8:15 am to 11:45 am or 1:15 pm to 4:45 pm. Of the enrolled children from 2 to 5
years, 44% of the AKA HS and EHS populations were immigrant and refugee
families consisting of the following demographics: Spanish (34%), Middle
Eastern and South Asian (8%), East Asian (1%), and African (1%). The AKA Early
Head Start program serves 214 children under three years of age of which 43% were
identified as immigrant or refugees—who were Spanish (37%), Middle Eastern and
South Asian (5%), and East Asian (1%) (AKA Head Start/Early Head Start Annual
Report 2014-2015; Appendix J).
A target of 800 to 1000 parents or caregivers who potentially met the
participant criteria were solicited to participate in the survey with the hopes
of obtaining the sample size requirement of 110 completed surveys. To
participate, a parent or caregiver had to be first-generation, foreign born,
and considered an immigrant or refugee. They must have been living in Southern
California and have children enrolled in either an Early Head Start or Head
Start program. A power analysis yielded an estimated sample size of 110 for a
linear regression with four predictors (power of 0.8, type I error of 0.05, and
medium size effect of 0.1) (Faul, Erdfelder, Buchner, & Lang, 2009);
however, all attempts were made to obtain the largest possible sample.
This study used a censuses sampling
procedure by attempting to reach all parents that met the criteria for inclusion
in the study. The sample was readily available to obtain in person at the 12
sites of AKA HS/EHS. Individuals were self-selected into the sample by choosing
to complete the survey (Fowler, 2009). The researcher asked the AKA Head Start
executive director and center directors to collaborate with the project to
ensure all immigrant/refugee parents/caregivers had a chance to participate. In
this way, the researcher avoided bias that could affects the relationship
between a sample of respondents and the population (Fowler, 2009). To avoid a
problem of bias, the researcher attempted to gather data from every member of
the sample population or sampling frame as the surveys were distributed at a
center’s gate where parents/caregivers would drop-off their children.
A self-administered paper-based
survey method was used for several reasons. First, the sample in this
population may not have had a computer or Internet access on a regular basis to
do an email survey. Second, the researcher was able to identify and access the
sample population with relative ease in person at AKA Head Start centers.
Third, the participants, the parents or caregivers, could read and interpret
survey questions in their own language and then answer with restricted
selection options, such as circling a number or checking a mark, which
eliminated the need for someone else to read the questions for respondents
(Fowler, 2009). Fourth, parents or caregivers were likely to cooperate with the
researcher by presenting their perceptions about parental involvement using
this method. Finally, some parents or caregivers were usually busy working and
had no time to do a survey or had a little time to be at home with their child.
This self-administered survey allowed them to complete the survey at their own
convenience (Fowler, 2009). The method of the self-administered survey was most
appropriate for this study; however, there were some disadvantages to this
approach. First, printing hundreds of paper surveys were cost the researcher
along with postages and self-addressed stamped envelopes. Second, returning the
surveys from respondents took a longer time than the three to four weeks
expected (Fowler, 2009). The researcher also needed additional time to enter
the responses into an electronic format (Blaxter, Hughes, & Tight, 2014), in
this case the Statistical Package for the Social Sciences (SPSS) 22.0.
Furthermore, tracking completed surveys and incentives cost the researcher’s
time (i.e., creating an Excel worksheet listing potential respondents who
completed the survey and whether an incentive was provided to them) (Survey
Administration Guidelines, 2009).
The Parent Involvement Project
(PIP) survey (Appendix A) was the instrument used for data collection.
Permission was obtained prior to using this instrument (Appendix I). The questionnaire
contains 57 items developed by Hoover-Dempsey, Walker, Jones, and Reed (2002).
All items were measured on a six-point Likert scale with subscales of attitudes
and beliefs, subjective norms, perceived behavioral control, and parental
intentions for involvement. The Likert response scale ranges from (1) strongly disagree to (6) strongly agree, with the additional
option of I Don’t Know. The options were: 1 indicating Strongly Disagree, 2 indicating Disagree,
3 indicating Don’t Know, 4 indicating
Agree Just a Little, 5 indicating Agree, and 6 indicating Strongly Agree. Parents were asked to
rate each statement based on how much they disagree
or agree. The survey takes
approximately 30 minutes to complete.
The Parental Involvement Project (PIP) questionnaire had established
reliability and validity (Hoover-Dempsey, Sandler, & Walker, 2005) and was
used in this study to assess key variables posed in the TPB model (i.e.,
parental attitudes and beliefs, subjective norms, and perceived behavioral
control) and ascertain whether they were significantly related to and could predict
the reported intentions of involvement of immigrant and refugee parents in
their children’s early childhood education programs. The survey had five
sections: (1) parental attitudes and beliefs, (2) subjective norms, (3)
perceived behavioral controls, (4) parent’s intention to become involved, and
(5) household demographics information.
The construct of parental attitudes
and beliefs is a learned predisposition
to respond to an object or class of objects in a consistently favorable or
unfavorable way (Kiriakidis, 2015). This variable was measured by using statements
that reflected parental attitudes and beliefs. The goal was to ascertain
individual’s behavioral attitudes and beliefs toward getting involved in
children’s education and the patterns of parental behavior that followed the
beliefs based on parents’ motivations for being involved (Hoover-Dempsey et
al., 2005). This subscale included 24
items (e.g., My child’s learning
is mainly up to the teacher and my child) with responses ranging from 1 = strongly agree to 6 = strongly
disagree. The subscale score for these items ranges from 24 to 144.
The
construct of subjective
norms is an individual’s perception of significant others’ beliefs; how much a
person is influenced by the judgment of significant others such as teachers,
parents, friends, or spouse (Ajzen, 1991). The subscale consisted of six statements constructs to reflect subjective
norms (e.g., I think most parents
at my child’s center are actively involved) with responses ranging from 1 = strongly agree to 6 = strongly
disagree. The subscale score for these items ranges from 6 to 36.
The
construct of perceived
behavioral controls (PBC) represents an individual’s perceived ease or
difficulty of performing a particular behavior (Ajzen, 1991). Any behavior is rarely under complete
volitional control and PBC can only be identified in relation to the
individual. Many external and internal factors could potentially inhibit the
intended execution of any behavior; therefore, the predictive role of PBC would
depend on the degree to which the behavior was under volitional control and the
potential role of external and internal factors to interfere with the behavior
(Kiriakidis, 2015). Thus, the greater the behavior depended on these factors
being enacted, the greater the predictive and explanatory role of PBC would be (Ajzen,
1991). These factors are assumed to reflect experience as well as
anticipated impediments and obstacles. The
subscale included 17 statements regarding parents’ ability to be involved (e.g.,
I have enough time and energy to attend
special events at school) with responses ranging from 1 = strongly
agree to 6 = strongly disagree.
The subscale score for these items ranges from 17 to 102.
The dependent variable of parental
intentions represents an indication of an individual parent’s readiness to
perform a given behavior, in this case parent involvement in schooling (Ajzen,
1991). Parental intentions are determined by attitudes and beliefs towards a
parental behavior, subjective norms or pressures, and perceived control to
perform the behavior and the parents’ motivation to comply (Kiriakidis, 2015). The subscale includes ten
statements (e.g., Your child’s
teacher asks you to schedule a conference to discuss your child’s progress)
with responses ranging from 1 = strongly agree to 6 = strongly disagree. The subscale score
for these items ranges from 10 to 60.
The final
section of the survey included household
demographic questions requesting participants to report their family’s general
information, such as their age, language spoken at home, and confirm their immigration/refugee
status. Demographic data was collected to ensure all participants were at least
18 years and up and met criteria to be eligible for the study. The survey questions
consistently reflected the construct that was measured (Field, 2013). The subscales
have high internal consistency, the reliability of the proportion of variance
attributable to the true score of the latent variable (Field, 2013). The
reliability scale utilized was the Cronbach alpha (α) because the survey
consisted of many Likert items (Trochim & Donnelly, 2008). The coefficient
alpha distinguished between the amount of variation, which stemmed from the
latent variable, and the amount attributable to error. The alpha coefficient ranges
in value from 0.0 to 1.0; however, when assessing the internal consistency, a
scale bellowed .60 is unacceptable; between .60 and .65 is undesirable; between
.65 and .70 is minimally acceptable; between .70 and .80 is respectable; and
between .80 and .90 is magnificent (Field, 2013). The reliability of the scale
was reported by Hoover-Dempsey, Sandler, and Walker (2002). The subscale
reliability of the 24 items measuring attitudes and beliefs was acceptable (α =
.77). The subjective norms subscale had a reported subscale reliability that was
respectable (α = .78). The subscale of perceived behavioral control consists of
17 items and had a subscale reliability reported as α = .83. Finally, the
parental intention subscale was composed of ten items with a reported subscale
reliablility of α = .78 (Hoover-Dempsey et al., 2005).
Field (2013) stated that an
instrument was valid when it measures what it set out to measure. For this
study, the concern was whether the survey accurately measured parental
involvement constructs of attitudes and beliefs, subjective norms, perceived
behavioral control, and intentions. Empirical work on developing the constructs
was included in the reports of Hoover-Dempsey and her colleagues (1995; 1997;
2002; 2005). They focused on the three factors of parents’ motivations and
considered them the most useful in the model of parental involvement: (1) an
active role construction of involvement (i.e., parents believe that being
involved in their children’s education wass important to their learning
development); (2) parental perceptions on being invited to be involved through
teachers, school or office staff creating a social norm to encourage parent
involvement; and (3) parents’ life context as critical. Parents’ understanding
of their own skills and knowledge influenced their thinking about the types of
involvement activities they took on. The parents’ perceptions on their
available time and energy for involvement also influenced their decisions. The
family culture played a significant role in their ideas about the ways they
could be and were involved in supporting their child’s learning (e.g., even
when children’s centers invite parents, their culture traditionally limited
parent or caregivers’ role in children’s formal schooling) (Hoover-Dempsey et
al., 2005). The design of this study included three independent variables and
one dependent variable which are operationally defined in the following
section.
Operational
Definitions of Variables
The
investigation included three independent variables considered as predictor
variables and a single dependent variable as an outcome variable. The three
independent variables were (a) attitudes and beliefs, (b) subjective norms, and
(c) perceived behavior control. The single dependent variable was the parent’s
reported intentions for involvement.
Each variable was operationally defined
as follows:
Attitudes/Beliefs. Parental
attitudes and beliefs were measured using 24 items with responses ranging from 1 = strongly agree to 6 = strongly
disagree. The individual items responses were summed for all items on the
subscale to create an interval level measured for this variable. The subscale
score for these items ranged from 24 to 144 with a low score indicating
somewhat negative attitudes and beliefs regarding parent involvement and a high
score indicating relatively positive attitudes and beliefs regarding parent
involvement.
Subjective norms. Subjective
norms were measured using six items constructed to reflect subjective norms. These responses ranged from 1 = strongly
agree to 6 = strongly disagree.
The individual item responses were summed for all items on the subscale to
create an interval level measured for this variable. The subscale score for this
variable ranged from 6 to 36 with a low score indicating somewhat negative
subjective norms regarding parent involvement and a high score indicating
relatively positive subjective norms regarding parent involvement.
Perceived behavioral controls.
Perceived behavioral controls were measured using 17 items regarding the
parents’ ability to get involved.
The responses ranged from 1 = strongly agree to 6 = strongly disagree. The individual item
responses were summed for all items on the subscale to create an interval level
measure for this variable. The subscale score for these items ranged from 17 to
102 with a low score indicating little perceived control regarding parent
involvement and a high score indicating considerable perceived control
regarding parent involvement.
Parental intentions. Parental
intentions was measured using ten items. The responses ranged
from 1 = strongly agree to 6 = strongly disagree. The individual item
responses were summed for all items on the subscale to create an interval level
measured for this variable. The subscale score for these items ranged from 10
to 60 with a low score indicating weak intentions to engage in parent
involvement and a high score indicating strong intentions to engage in parent
involvement.
Study Procedures
The Northcentral University
Institutional Review Board (IRB) approved the research. The AKA Head Start
executive director and all center directors also granted access and approval to
conduct the study at their sites. The research materials and instruments were
sent to all gatekeepers at the same time within an email. Each contained (a) a
letter of asking permission to conduct the surveys (Appendix E), (b) a cover
letter of an introduction to the research project and information about the
nature and purpose of the study (Appendix D), and (c) a copy of the survey (Appendix
A). The researcher met with the Head Start executive director and each of the
12 center directors, as first and second gatekeepers, to discuss whether the
research inquiry could proceed and the process of how to contact
parents/caregivers and distributed the surveys they would prefer. The final
plan agreed upon for times of soliciting the participants was during the drop-off
and pick-up times at each location. Once all approvals were received, the distribution
of parent surveys began with a pre-notice letter (Appendix F). This pre-notice
letter was put into each child’s mailbox at all 12 Head Start/Early Head Start
centers for parents/caregivers. A few days after passing out the pre-notice
letters, during the drop-off time at each of twelve centers, the researcher met
parents or caregivers and briefly explained the purpose of the study and allowed
them to ask any questions and asked if they would be interested in cooperating
with the research voluntarily (Fowler, 2009). Those interested were given a
survey package with (1) a survey, (2) a consent form (Appendix B), (3) a recruitment
letter (Appendix G), and (4) a self-stamped envelope for participants to return
the survey and consent signature to the researcher’s address if they prefer
this mode. All materials were translated into their home language (i.e.,
Arabic, Spanish, and Vietnamese) to ensure parents/caregivers comprehended
explanations thoroughly.
The recruitment letter communicated
(a) the purpose of this study; (b) a request for their voluntary participation
with no repercussions for not participating; (c) the importance and usefulness
of participation, and (d) requirement of an informed consent document; (e)
information of the risk level of involvement and absolute no-deception; (f)
respondents’ names and information remaining confidential (i.e., names of
respondents are not associated with the results in any reporting) and anonymous
(i.e., names of respondents are not known); (g) the name and contact
information of the research supervisor for this project at the Northcentral
University (NCU) for further questions regarding the survey that respondents could
use to inquire; and (h) a timeline for completing and returning the survey with
an offered incentive for the survey respondents (Fowler, 2009). The researcher offered
a gift card if the survey was completed on the same day or the participants took
the surveys home and returned them within two to three weeks.
The next contact was a reminder
postcard placed in each child’s mailbox that encouraged the participants who had
not completed and returned the survey to do so if they were still interested in
participating (Blaxter, Hughes, & Tight, 2014). The final contact was a thank-you
letter for respondents’ time and consideration, accompanied with a gift card of
ten dollars for those who returned the survey materials within the timeline. A gift
card and a thank-you letter were sent to their home address by post office.
Data Collection and
Analysis
All surveys were completed and
collected within an eight-week timeframe. The survey results were entered into
a spreadsheet. The data from the spreadsheet were entered into the Statistical
Package for the Social Sciences (SPSS), version 22 for final statistical analysis.
A power analysis yielded an estimated sample size of 110 for a linear
regression with three predictors (power of 0.8, type I error of 0.05, and
medium size effect of 0.1) (Faul, Erdfelder, Buchner, & Lang, 2009);
however, the researcher would had a goal of collecting data from at least 500
participants to account for missing data or incomplete surveys. The goal was to
achieve a response rates between 30% and 40% (Fowler, 2009). A missing data
rate of 15% to 20% is common in educational studies (Dong & Peng, 2013).
Dong and Peng (2013) found that 36% of studies had no missing data, 48% had
missing data, and about 16% could not be determined. Missing data reduces the
statistical power of a trial and impacts to the quality of statistical
inferences, which refers to the probability that would reject the null
hypothesis (Dong & Peng, 2013; Kang, 2013). The missing rate of 5% or less
was inconsequential; however, if there was more than this rate, then the bias
of statistical analysis could likely happen (Dong & Peng, 2013). Moreover,
the impact of missing data on quantitative research could be serious because there
is loss of information, decreased statistical power, increased standard errors,
and weakened generalizability of findings (Kang, 2013). During the survey
process, missing data can be caused by several factors: (1) respondents refused
or forgot to answer a question because of privacy issues, (2) the person taking
the survey did not understand the question due to a lack of experience or
reading skill, (3) respondents lost interest or did not have enough time to
complete the questionnaire, (4) respondents did not show up on the survey day,
and (5) databases had missing data because there was a mismatch of variables
between databases.
The data was screened for accuracy
to clarify any problems or errors (Trochim & Donnelly, 2008). Missing data
was handled by applying the multiple imputation approach (Trochim &
Donnelly, 2008). In this approach, replacing with a set of plausible values was
the strategy for the missing values in which they contain natural variability
and uncertainty of right values (Kang, 2013). The first step before a data set
with missing values is analyzed by statistical procedures is that it needs to
be edited in some ways into a complete data set. Failure to edit the data
properly could make it unsuitable for a statistical procedure and the statistical
analyses were vulnerable to violations of assumptions (Dong & Peng, 2013).
Second, a prediction of the missing data was completed by using the existing
data. Then the missing values were replaced with the predicted values, and a
full data set called the imputed data set was created. This process iterates
the repeatability and creates multiple imputed data sets (Kang, 2013). Kang
stressed each multiple imputed data set produced be analyzed using the standard
statistical analysis procedures for complete data, thereby giving multiple
analysis results. By combining these analysis results, a single overall
analysis result was produced. In addition to restoring the natural variability
of the missing values, it incorporated the uncertainty due to the missing data,
which results in a valid statistical inference (Kang, 2013). Moreover,
restoring the natural variability of the missing data could be achieved by
replacing the missing data with the imputed values using the regression method,
or the predictive mean matching method ccould be used if the missing variables were
continuous (Dong & Peng, 2013). Furthermore, multiple imputations are robust
to the violation of the normality assumption and produces appropriate results
even in the presence of a small sample size or a high number of missing data
(Dong & Peng, 2013). SPSS has a missing value analysis module that allowed
examination of the patterns of data completion in a descriptive way. This test identified
whether the data significantly departed from missing data completely at random
(Trochim & Donnelly, 2008). In general, the mean, standard deviation, and
frequencies were checked on the amount of missing data, and the regression
method was utilized to impute the missing values. Handling the missing data by
using the SPSS Missing Value Analysis was helpful in solving any missing data
problems (Kang, 2013). Cronbach’s alphas were used to compute and determine the
internal consistency reliability of each subscale (Trochim & Donnelly,
2008).
Descriptive summary statistics,
including the calculations of means, standard deviations, and ranges for
attitudes and beliefs, subjective norms, perceived behavior control, and
parental intentions were obtained to answer the first three research questions.
Because this study had one dependent variable and multiple independent
variables, a multiple regression analysis was used to determine the amount of
variation in the dependent variable explained by the independent variables to
answer research question four. The step-wise multiple regression employed had
the following assumptions that must be met: linearity, collinearity,
independence of errors, normality, and homoscedasticity. Statistical SPSS
software was used to test each assumption. These assumptions were tested through
a visual inspection of data plots, skewness, kurtosis, Q-plots, P-plots, and
VIF statistics (Field, 2013). Skewness and kurtosis were checked in the statistical
tables and normality was checked through histograms and plots of the
standardized residuals (Field, 2013). If the assumption of normality was not possible
reasons why were investigated (e.g., the underlying distribution was nonnormal,
outliers or mixed distribution of scores on each of the variables might
contain, a low discrimination gauge was used, skewness was present in the data)
(Foster, Barkus, & Yavorsky, 2006) and then a corrective procedure was
performed, such as transformations if needed. If the assumption of linearity
was not met, then the item correlation matrix was examined to identify any item
that did not correlated with the subscale and eliminate such items was
considered with subscale reliabilities being recomputed (Foster, Barkus, &
Yavorsky, 2006). If the assumption of no multicollinearity was not met, then
the variables with a low value on tolerance would be removed. If the assumption
of homoscedasticity was not met, then a transformation of the variables or use
of a weighed least squares regression was considered (Foster, Barkus, &
Yavorsky, 2006).
Summary statistics of demographics
regarding ages and genders were computed and reported to describe the sample.
To answer research question four, the SPSS analysis resulted in three main
tables: the Model Summary table that reported the R and R^2 and adjusted R^2 and standard error to indicate how
well the data fit the model, the ANOVA table that reported the F-ratio and significance for the overall
regression model, and the Coefficients table that identified the coefficients
for each independent variable and if it was a significant variable in the
regression model (Laerd Statistics, n.d.).
The assumptions in this study included:
(a) the application of the theory of
planned behavior model was appropriate to be utilized in predicting/explaining
and gaining a deeper understanding of immigrant and refugee parents/caregivers’
intentions with regard to parental involvement, (b) participants would
answer all survey questions truthfully and honestly, (c) the variables of
parental attitudes and beliefs, subjective norms, perceived behavioral control,
and parental intention regarding parent involvement were complex and would be
fully measured by the survey, (d) the survey instrument used was valid
and reliable, and (e) the sampling was unbiased. These assumptions had a
potential effect on the characteristics of data, such as distribution trends,
correlational trends, and variable included.
Limitations
The following limitations were
present in this study. A possible limitation for the immigrant and refugee parent
participants was a language barrier. To compensate for this limitation, the
survey documents were provided in English and in their primary language. Also,
since the surveys were sent home, participants had no opportunity to clarify
their confusion, if there was any, thereby possibly inaccurately interpreting one
or more questions. Also, parental reading levels could impact comprehension of
each question or prevent parents/caregivers from volunteering to take the
survey (Keys, 2015). The researcher’s email address
and phone numbers were provided for participants in case they had inquiries
that arose regarding the survey during the study. Because the scope of
this study included only a sample of immigrant/refugee families, the findings
are limited to the sample obtained from those enrolled in Head Start and Early
Head Start programs in the vicinity. This study was also limited by the use of a convenience sample. These limitations increased the
possibility of common-method bias, which increased the probability that the
characteristics of those who responded were different from those who did not.
Alternatively, anonymity provided through anonymous survey helped counteract
some biases vesus a focus groups of immigrants and refugees that would be
susceptible to cultural correctness.
Delimitations
The
problem of parental involvement in children’s education in Head Start or Early
Head Start programs was the focus of the study. Although there are other
problems within immigrant/refugee families, the findings were delimited to only
those variables being measured in relation to parent involvement and the theory
of planned behavior model. The criteria for participants’ enrollment in this
study were first-generation, foreign-born immigrant and refugee parents, which eliminated
some parents who were not qualified to participate even though they would
consider themselves a part of this population. This study was delimited to those
who lived in a geographic regions in Southern California and whose children
were enrolled in Head Start and Early Head Start programs.
Regarding
parental behavior, Ajzen’s theory of planned behavior (TPB) model (1991) was a
useful method to explain and predict parents’ intentional behavior based on
their personal beliefs about the outcomes of behaviors. The TPB was applied to describe
the dynamic and complex nature of parental engagement in their child’s life and
education (Bracke & Corts, 2012). In addition, the TPB based moldel offered
a viable theoretical lens for examining parental involvement, the most
important determinant of parental behavioral dispositions. However, based on
these goals, the theory of planned behavior model may not captured all aspects
of parents/caregivers and their intentions for parent involvement, yet the
study delimited to only the three independent variables included in the model. These
delimitations had a potential effect on the examination of relationships among beliefs/attitudes,
subjective norms, perceived behavioral controls, and parental intentions. Other
variables may contribute to parent involvement or a lack of involvement that
were not measured or were not accounted for in the study; therefore, the
prediction formula is limited.
Given
the study included human participants, ethical assurances were required. There
were minimal elements of risk, and no deception was used in this study. Other
assurances included privacy, informed consent, anonymity, secrecy, being
truthful, and confidentiality (Moreno, Goniu, Moreno, & Diekema, 2013). Federal
regulations defined a human subject as a living individual about whom an
investigator obtained data through interaction with the individual or
identifiable private information (Moreno, Goniu, Moreno, & Diekema, 2013).
Informed consent, confidentiality, and protection of individuals were central
to the guidelines on research ethics and were employed in the study (Blaxter, Hughes,
& Tight, 2014). There was a statement in the recruitment letter for consent
to participate; and the researcher required a signed informed consent form
before collecting data from parents/caregivers and the consent forms and
information were kept confidential and separate from data after collection
(Moreno, Goniu, Moreno, & Diekema, 2013). The researcher’s assurances and
trust included the absence of deception, the voluntary nature of participation,
and the risk involved.
Further, it was critical to protect
the participant identities (Fowler, 2009). Their names, email addresses, postal
address and telephone numbers did not appear on the survey. Data and results
from the survey did not include personal information, and the surveys were not shared
with anyone other than the researcher and the committee at Northcentral
University. Concerning participants’ information on all documents such as a
consent form and demographic information, they will be destroyed after seven
years of being kept in the researcher’s personal locked files (Fowler, 2009).
The collected information was utilized only in support of this study and only
for educational purposes. The final reports of the study contain no personal
information of parents’/caregivers, maintaining their confidentiality.
Summary
To apply
the theory of planned behavior model to predict/explain and gain a deeper
understanding of the immigrant and refugee parents/caregivers’ intentions and
behaviors with regard to parental involvement, it was necessary to utilize a
quantitative method and a cross-sectional survey design. A descriptive and correlational
design was utilized to identify the parents’
intention and behaviors and to answer the research questions. Upon receiving
immigrant and refugee parents’ completed surveys, their responses were
inputted into SPSS. The instrument used to collect data from parents or caregivers
had four sections measuring perceptions of parents or caregivers: parental
attitudes/beliefs, subjective norms, perceived behavioral controls, and
parental intentions. The instrument had sufficient reliability and validity.
Descriptive statistics and inferential data analysis were used within SPSS to
answer the research questions. The analyses included the calculations of means,
standard deviations, frequencies, and percentages of parental intentions of
involvement. Subscale reliabilities and correlations were calculated. A
stepwise multiple regression was used to answer the final research question.
No comments:
Post a Comment