3.1 Re-coding to an Ordered Factor: Education

The variable EDUC is a code between 000 and 125 for the level of education, and includes values of 999 for missing or unknown values. The following screenshots from https://cps.ipums.org/cps-action/variables/EDUC#codes_section show the meaning for each code.

EDUC Page	Codes
-	-

Let us start by creating a new variable in our data, called edu, which will become an ordered factor with descriptive levels. We initially set all values to the missing value, NA

df$edu <- NA

Next, let us define the following levels. The following code just creates a short list of strings that will become the levels for df$edu.

edulevels = c("Less than high school",
              "High school degree or equivalent",
              "Some College",
              "Four Year College Degree",
              "Advanced College Degree")

Next, let us assign the observations for the new variable df$edu based on the codes in df$EDUC.

Those with EDUC values greater than 0 (exactly equal to 0 is “not in universe”, which we will keep as NA, i.e. as missing) and less than 73 have less than a high school education. This description is stored in our first element of edulevels:

df$edu[ df$EDUC>0 & df$EDUC<73 ] <- edulevels[1]

Next, those with EDUC exactly equal to 73 have a high school diploma or equivalent, the second element of edulevels:

df$edu[ df$EDUC==73 ] <- edulevels[2]

These with EDUC values greater than 73 and less than 111 have some college, but not a bachelor degree. This is associated with the third element of edulevels, “Some College”.

df$edu[ df$EDUC>73 & df$EDUC<111] <- edulevels[3]

Those with EDUC values greater than or equal to 111 and less than 123 have a bachelor degree, but not an advanced degree. We associate this with our fourth level, “Four Year College Degree”.

df$edu[ df$EDUC>=111 & df$EDUC<123] <- edulevels[4]

Finally, those with EDUC values greater than or equal to 123 have some sort of advanced degree. We associate this with our final level.

df$edu[ df$EDUC>=123] <- edulevels[5]

Now, we transform the variable df$edu to an ordered factor, because it is a categorical variable whose levels have a meaningful order. In the code below, we overwrite df$edu, using its own data, specifying the levels the edulevels, and specifying that it is ordered.

df$edu <- factor(df$edu, levels=edulevels, ordered=TRUE)

We can now view the class, levels, and summary of our new variable.

class(df$edu)

## [1] "ordered" "factor"

levels(df$edu)

## [1] "Less than high school"            "High school degree or equivalent"
## [3] "Some College"                     "Four Year College Degree"        
## [5] "Advanced College Degree"

summary(df$edu)

##            Less than high school High school degree or equivalent 
##                            66103                            39419 
##                     Some College         Four Year College Degree 
##                            38705                            26476 
##          Advanced College Degree 
##                            15211

The output to class(df$edu) confirms that we have created an ordered factor. The output levels(df$edu) shows that our levels for education have been established. Finally, the output of summary(df$edu) gives us some frequency statistics regarding our sample.

3.2 Re-coding a Factor: Worker Classification

The variable, CLASSWKR, is a code between 10 and 29 for the type of worker. The following screenshot from https://cps.ipums.org/cps-action/variables/CLASSWKR#codes_section shows the meaning for each code.

We first create a new variable in our data frame and call is class. We initially set all values equal to the missing observation code NA.

df$class <- NA

Those with codes equal to 10 through 14 are self employed. Here we create a category called “Self-Employed” for this group and assign this label to the appropriate observations:

df$class[ df$CLASSWKR>=10 & df$CLASSWKR<=14 ] <- "Self-employed"

Those with a value for CLASSWKR exactly equal to 21 are employed by private companies. Here we create this category and assign this label to the appropriate observations:

df$class[ df$CLASSWKR==21 ] <- "Employed by private company"

Those with a value for CLASSWKR between 24 and 28 have government employers. Here we create this category and assign the label to the appropriate observations:

df$class[ df$CLASSWKR>=24 & df$CLASSWKR<28] <- "Employed in Government"

Finally, those with a value for CLASSWKR exactly equal to 29 are unpaid family workers. Here we create this last category and assign the label to the appropriate observations:

df$class[ df$CLASSWKR==29] <- "Unpaid Family Worker"

Finally, we will cast our new variable class as a factor variable (i.e. a categorical variable):

df$class <- factor(df$class)

3.3 Coding Missing Observations in Numerical Data

Usual hours worked per week (UHRSWORKT) and weekly earnings (EARNWEEK) are both numerical variables, but they both have numeric codes for missing values.

The following screenshot from IPUMS for the usual hours worked per week variable (https://cps.ipums.org/cps-action/variables/UHRSWORKT#codes_section) shows that the code is numeric, but uses codes 997 for “hours vary” and 999 for “Not in universe (NIU)”, i.e. a missing value.

In the code below, we create a new variable called hrsworked, set it equal to UHRSWORKT. Then, where observations for UHRSWORKED are equal to 997 or greater, we set hrsworked equal to the missing value code, NA.

df$hrsworked <- df$UHRSWORKT
df$hrsworked[ df$UHRSWORKT>=997] <- NA

The following screenshot from IPUMS for the weekly earnings variable (https://cps.ipums.org/cps-action/variables/EARNWEEK#codes_section) shows that the code is numeric, but uses code 9999.99 as “Not in Universe,” i.e. a missing value.

In the code below, we create a new variable called earnweek, set it equal to EARNWEEK. Then, where observations for EARNWEEK are equal to 9999.99, we set earnweek equal to the missing value code, NA.

df$earnweek <- df$EARNWEEK
df$earnweek[ df$EARNWEEK == 9999.99 ] <- NA