San Diego State University Variable Biostatistics Questions

Selected research questionHow do 12th graders’ frequency of missing school due to illness differ by sex among students over
18 and students under 18?
Key variables used in our analyses
The variables included in our analysis were: a) Sex (V2150: 00030:R’S SEX) renamed
Sex_Gender, b) Age (Item #: 89848: Age18Dichotomy) renamed Age and c) missing school
illness (V2175: 00430: #DA/4W SC MS ILL. Renamed: SchoolDays_MS_illness_last4W_Di
a) The survey question for variable Sex_Gender (nominal) asks participants ‘What is your
sex?’. The optional categories Male and Female. The dataset recorded those categories in
the dataset so that 1= ‘Male’, 2= ‘Female’
b) The survey questions for Age (ordinal) included “In what year were you born?”, “In what
month were you born?”, and the date of questionnaire administration as recorded by the
interviewer. Values for the variable included 1= ‘under 18 years old’ and 2= ‘18 years of
age and over’.
c) We only recoded one variable in our analysis into a new variable. The survey question for
variable SchoolDays_MS_illness_last4W (Ordinal) asks participants ‘During the LAST
FOUR WEEKS, how many whole days of school have you missed…because of illness?’.
Values for the variable include 1= ’None’, 2= ‘1 DAY’, 3= ‘2 DAYS’, 4= ‘3 DAYS’, 5= ‘45DAYS’, 6= ‘6-10 DA’, 7= ‘11+DAYS’, and 9= ‘MISSING’. Our new variable is
SchoolDays_MS_illness_last4W _Di using the value 1= ‘None’ changed to 1= ‘NO
SCHOOL MISSED LAST4WEEKS, 2= ‘1 DAY’, 3= ‘2 DAYS’, 4= ‘3 DAYS’, 5= ‘45DAYS’, 6= ‘6-10 DA’, 7= ‘11+DAYS’’ changed to 2= MISSED SCHOOL (1 Day+)
ILLNESS. We decided to recode the SchoolDays_MS_illness_last4W variable into
SchoolDays_MS_illness_last4W_Di because it was easier to see the distinction between two
categories as opposed to several. Missing variables remained as missing variables.
Characteristics of Each Variable
Table 1.1 (Sex)
Table 1.2 ( Age)
Table 1.3 (Missed School due to Illness)
Approach to Analysis
For our analysis we chose to run a cross tabulation (“crosstab”) in order to assess whether 12th
graders frequency of missing school due to illness differ by sex among students over 18 and
students under 18? The cross tabulation helps us see the frequency and proportions between two or
more variables. It will allow us to graph, identify patterns in relationship, and avoid incorrect
interpretation when looking at one variable at the time. We will be using the crosstab to look at
gender/sex categories missed school due to illness in the last 4 weeks and who are over 18 and
under 18 years of age. We also ran a chi-square test statistically see if a relationship between two or
more categorical variables exists among 12th graders’ frequency of missing school due to illness
differ by sex among students over 18 and students under 18.
Our hypotheses for this test are as follows:
For students under 18 years of age:
● Null hypothesis (𝐻0 ): There is no relationship between 12th graders’ missing school due to
illness and sex/gender among students under 18
● Research/alternative hypothesis (𝐻1 ): There is a relationship between 12th graders’ missing
school due to illness and sex/gender among students under 18
For students 18 years of age and older:
● Null hypothesis (𝐻0 ): There is no relationship between 12th graders’ missing school due to
illness and sex/gender, among students over 18
● Research/alternative hypothesis (𝐻1 ): There is a relationship between 12th graders’ missing
school due to illness and sex/gender, among students over 18
To do this, we used the “crosstabs” feature in SPSS. We selected
SchoolDays_MS_illness_last4W_Di as our “row” variable, Sex_Gender as our “column” variable,
and Age as our “layer” variable. We chose age as our layer because we want to stratify our results
by age in order to see the differences in missed days by sex_gender for those students who are
under and over 18 years of age. We also asked SPSS to produce a Chi-Square test statistic and pvalue within the crosstabs feature (under the “statistics” tab).
Table 2 (Crosstab: Column Percentage)(Row: Missed School)(Column: Sex/Gender)(Layer:
Age)
Chi-Square Test Table 3.0 (Goodness of fit Test)
Describe Implications and limitations (Chrestina Askoore )Look at lecture 5 slides 31-32
EXAMPLE: Our findings suggest that there is a statistically significant relationship between
race/ethnicity and number of hours worked per week during the school year among students under
18 years old as well as among students over 18 years old. We found that results were similar
regardless of age (over or under 18) which means that we may not have gained any additional
information or nuance by stratifying our results by age. In the real world, we might re-run our
analyses without the “layer” (i.e., stratifying) variable because it is not needed to help tell the story
of our data.
Our findings begin to tell a story between race/ethnicity and working during the school year, which
may have implications for 12th graders’ well-being. For example, if students who are working more
tend to be one racial/ethnic group versus another, this could result in inequitable well-being among
the group(s) that work more. However, we do not know why one group may be working more than
others and there are activities that might take up students’ time, for example caring for family
members, that would not be captured in the variable about work, but still may affect well-being.
Future research is needed to understand how race/ethnicity and number of hours worked per week
during the school year may impact the well-being of 12th graders in the U.S.
Limitations
EXAMPLE: As mentioned in our results section, the measurement of our variables were restrictive.
Specifically, our rec-coding of the original work hours variable — which originally had multiple
categories — may have missed important differences by number of hours worked when we changed
it to “no work” vs. “1 or more hours of work”. By making this a dichotomous variable, we may
have over-simplified our analyses and as such, masked important information. Indeed, students who
work one hour per week during the school year likely have a very different experience with work
and task-management than those students who work 5, 10, or more hours per week. Our results are
further constrained because we do not know more details about age or race/ethnicity in our
analyses. For example, we know that there are more than 3 racial/ethnic categories among people in
the United States, but the survey administrator only provided us with data on three of the largest
groups. Similarly, we are unable to examine differences by more categories of age beyond under 18
versus over 18.
An important limitation of our analyses is our use of the 2019 Monitoring the Future Survey for
12th graders only. This is a cross-sectional dataset which does not allow us to examine changes in
individuals over time and we are only looking at outcomes for 12th graders, even though other
grades do participate in the survey. Had we included the other grades, our data might have told a
fuller picture about the relationship between race/ethnicity, work hours per week during the school
year, and age. Finally, our analyses are only descriptive in nature. They do not give us insights
about how or why different variables are related to one another. All we can tell is that our chisquare test of independence showed a significant relationship between race/ethnicity and hours
worked, regardless of whether 12th graders are over or under 18. .