This module shows how to create and recode variables. In SPSS you can create new variables with compute and you can modify the values of an existing variable with recode.
1. Computing new variables
Let’s use the auto data for our examples. In this section we will see how to create new variables with compute.
get file 'c:https://stats.idre.ucla.edu/wp-content/uploads/2016/02/auto.sav'.
The variable length contains the length of the car in inches. Below we see summary statistics for length.
descriptives variables = length.
Descriptive Statistics N Minimum Maximum Mean Std. Deviation Length (in.) 74 142 233 187.93 22.266 Valid N (listwise) 74
Let's use the compute command to make a new variable that has the length in feet instead of inches, called lenft.
compute lenft = length / 12. execute. descriptive variables=length lenft.
Descriptive Statistics N Minimum Maximum Mean Std. Deviation Length (in.) 74 142 233 187.93 22.266 LENFT 74 11.83 19.42 15.6610 1.85553 Valid N (listwise) 74
Suppose we wanted to make a variable called length2 which has length squared.
compute length2 = length**2. execute. descriptive variables = length2.
Descriptive Statistics N Minimum Maximum Mean Std. Deviation LENGTH2 74 20164.00 54289.00 35807.6892 8364.04524 Valid N (listwise) 74
Or we might want to make loglen which is the natural log of length. Note that you can shorten the command descriptive to just desc, and you can shorten variables to var.
compute loglen = ln(length). execute. desc var = loglen.
Descriptive Statistics N Minimum Maximum Mean Std. Deviation LOGLEN 74 4.96 5.45 5.2290 .12014 Valid N (listwise) 74
Let's get the mean and standard deviation of length and we can make Z-scores of length. In SPSS there are two ways to get the z-scores, and we will show you both ways. The first way is to use the save subcommand after the descriptive command. This will save the z-scores into the data file. The other way to obtain z-scores is to make them manually, and the code necessary to do that is shown below. When making z-scores manually, you do not need to use the save subcommand with the descriptive command.
desc variables = length /save.
Descriptive Statistics N Minimum Maximum Mean Std. Deviation Length (in.) 74 142 233 187.93 22.266 Valid N (listwise) 74
The mean is 187.93 and the standard deviation is 22.27, so zlength can be computed as shown below.
compute zlen = (length - 187.93) / 22.27. execute. desc variables = zlen.
Descriptive Statistics N Minimum Maximum Mean Std. Deviation ZLEN 74 -2.06 2.02 .0001 .99984 Valid N (listwise) 74
With compute
you can use + - for addition and subtraction
you can use * / for multiplication and division
you can use
** for exponents (e.g., length**2)
you can use ( ) for
controlling order of operations.
2. Recoding new variables
Suppose that we wanted to break mpg down into three categories. Let's look at a table of mpg to see where we might draw the lines for such categories.
frequencies variables = mpg.
Statistics
Mileage (mpg)N Valid 74 Missing 0
Mileage (mpg) Frequency Percent Valid Percent Cumulative Percent Valid 12 2 2.7 2.7 2.7 14 6 8.1 8.1 10.8 15 2 2.7 2.7 13.5 16 4 5.4 5.4 18.9 17 4 5.4 5.4 24.3 18 9 12.2 12.2 36.5 19 8 10.8 10.8 47.3 20 3 4.1 4.1 51.4 21 5 6.8 6.8 58.1 22 5 6.8 6.8 64.9 23 3 4.1 4.1 68.9 24 4 5.4 5.4 74.3 25 5 6.8 6.8 81.1 26 3 4.1 4.1 85.1 28 3 4.1 4.1 89.2 29 1 1.4 1.4 90.5 30 2 2.7 2.7 93.2 31 1 1.4 1.4 94.6 34 1 1.4 1.4 95.9 35 2 2.7 2.7 98.6 41 1 1.4 1.4 100.0 Total 74 100.0 100.0
Let's convert mpg into three categories to help make this more readable. Here we convert mpg into three categories using compute and if.
compute mpg3 = 1. if (mpg >= 19) & (mpg <= 23) mpg3 = 2. if (mpg >= 24) & (mpg <= 100) mpg3 = 3. execute.
Now, we could use mpg3 to show a crosstab of mpg3 by foreign to contrast the mileage of the foreign and domestic cars.
crosstabs /tables = mpg by mpg3.
Case Processing Summary Cases Valid Missing Total N Percent N Percent N Percent Mileage (mpg) * MPG3 74 100.0% 0 .0% 74 100.0%
Mileage (mpg) * MPG3 Crosstabulation
CountMPG3 Total 1.00 2.00 3.00
Mileage (mpg) 12 2 2 14 6 6 15 2 2 16 4 4 17 4 4 18 9 9 19 8 8 20 3 3 21 5 5 22 5 5 23 3 3 24 4 4 25 5 5 26 3 3 28 3 3 29 1 1 30 2 2 31 1 1 34 1 1 35 2 2 41 1 1 Total 27 24 23 74
crosstabs /tables = mpg3 by foreign /cells = count column.
Case Processing Summary
Cases Valid Missing Total N Percent N Percent N Percent MPG3 * Car type 74 100.0% 0 .0% 74 100.0%
MPG3 * Car type Crosstabulation
Car type Total Domestic Foreign
MPG3 1.00 Count 22 5 27 % within Car type 42.3% 22.7% 36.5% 2.00 Count 19 5 24 % within Car type 36.5% 22.7% 32.4% 3.00 Count 11 12 23 % within Car type 21.2% 54.5% 31.1% Total Count 52 22 74 % within Car type 100.0% 100.0% 100.0%
The crosstab above shows that 21% of the domestic cars fall into the high category, while 55% of the foreign cars fit into this category.
3. Recoding variables using recode
There is an easier way to recode mpg to three categories using recode. Using this method, we do not need to make a copy of mpg or use the compute command. We simply use the recode command with the into option with the name of the new variable into which we want to recode mpg. In this case, we will recode mpg into mpg3a using three categories: lo-18 into 1, 12-23 into 2, and 24-hi into 3. Note the lo and hi are SPSS keywords that can be used when we do not know the lowest or the highest values of the variable.
recode mpg (lo thru 18=1) (19 thru 23=2) (24 thru hi=3) into mpg3a. execute.
Let's double check to see that this worked correctly. We see that it worked perfectly.
crosstabs /tables = mpg by mpg3a.
Case Processing Summary Cases Valid Missing Total N Percent N Percent N Percent Mileage (mpg) * MPG3A 74 100.0% 0 .0% 74 100.0%
Mileage (mpg) * MPG3A Crosstabulation
CountMPG3A Total 1.00 2.00 3.00
Mileage (mpg) 12 2 2 14 6 6 15 2 2 16 4 4 17 4 4 18 9 9 19 8 8 20 3 3 21 5 5 22 5 5 23 3 3 24 4 4 25 5 5 26 3 3 28 3 3 29 1 1 30 2 2 31 1 1 34 1 1 35 2 2 41 1 1 Total 27 24 23 74
4. Recodes with if
Let's create a variable called mpgfd that assesses the mileage of the cars with respect to their origin. This variable, mpgfd, will have two values:
0 if below the median mpg for its group (foreign/domestic)
1 if at/above the median mpg for its group (foreign/domestic).
sort cases by foreign.examine variables = mpg by foreign /plot none /compare group / percentiles (5,10,25,50,75,95) haverage.
Case Processing Summary Cases Valid Missing Total N Percent N Percent N Percent Mileage (mpg) 74 100.0% 0 .0% 74 100.0%
Descriptives Statistic Std. Error Mileage (mpg) Mean 21.30 .673 95% Confidence Interval for Mean Lower Bound 19.96 Upper Bound 22.64 5% Trimmed Mean 20.92 Median 20.00 Variance 33.472 Std. Deviation 5.786 Minimum 12 Maximum 41 Range 29 Interquartile Range 7.25 Skewness .968 .279 Kurtosis 1.130 .552
Percentiles Percentiles 5 10 25 50 75 95 Weighted Average(Definition 1) Mileage (mpg) 14.00 14.00 17.75 20.00 25.00 34.25 Tukey's Hinges Mileage (mpg) 18.00 20.00 25.00
Case Processing Summary Cases Valid Missing Total Car type N Percent N Percent N Percent Mileage (mpg) Domestic 52 100.0% 0 .0% 52 100.0% Foreign 22 100.0% 0 .0% 22 100.0%
Descriptives Car type Statistic Std. Error Mileage (mpg) Domestic Mean 19.83 .658 95% Confidence Interval for Mean Lower Bound 18.51 Upper Bound 21.15 5% Trimmed Mean 19.60 Median 19.00 Variance 22.499 Std. Deviation 4.743 Minimum 12 Maximum 34 Range 22 Interquartile Range 5.75 Skewness .794 .330 Kurtosis .612 .650 Foreign Mean 24.77 1.410 95% Confidence Interval for Mean Lower Bound 21.84 Upper Bound 27.70 5% Trimmed Mean 24.48 Median 24.50 Variance 43.708 Std. Deviation 6.611 Minimum 14 Maximum 41 Range 27 Interquartile Range 8.25 Skewness .706 .491 Kurtosis .468 .953
Percentiles Percentiles Car type
5 10 25 50 75 95 Weighted Average(Definition 1) Mileage (mpg) Domestic 13.30 14.00 16.25 19.00 22.00 29.35 Foreign 14.45 17.00 20.25 24.50 28.50 40.10 Tukey's Hinges Mileage (mpg) Domestic 16.50 19.00 22.00 Foreign 21.00 24.50 28.00
We see that the median is 19.00 for the domestic (foreign=0) cars and 24.50 for the foreign (foreign=1) cars. The compute and recode commands below recode mpg into mpgfd based on the median for the domestic cars and the median for the foreign cars. In this example, we show how to create a new variable with all missing values, which can then be recoded. In SPSS, to create a new variable with all missing values, you use the compute command and set the new variable equal to $sysmis. The SPSS system variable $sysmis creates system missing values. We also use the do if command, which is useful when you want to recode a variable based on different values of another variable. Remember that you will need to use an end if command at the end of your do-loop.
compute mpgfd = $sysmis. do if foreign = 0. recode mpg (lo thru 18=0) (19 thru hi=1) into mpgfd. end if. do if foreign = 1. recode mpg (lo thru 24=0) (25 thru hi=1) into mpgfd. end if. execute.
We can check the new variables using the command below. The recoded variable mpgfd looks correct.
crosstabs /tables = mpg by mpgfd.
Case Processing Summary Cases Valid Missing Total N Percent N Percent N Percent Mileage (mpg) * MPGFD 74 100.0% 0 .0% 74 100.0%
Mileage (mpg) * MPGFD Crosstabulation
CountMPGFD Total .00 1.00
Mileage (mpg) 12 2 2 14 6 6 15 2 2 16 4 4 17 4 4 18 9 9 19 8 8 20 3 3 21 2 3 5 22 5 5 23 3 3 24 1 3 4 25 5 5 26 3 3 28 3 3 29 1 1 30 2 2 31 1 1 34 1 1 35 2 2 41 1 1 Total 33 41 74
Summary
Create a new variable len_ft which is length divided by 12.
compute len_ft = length / 12.
Recode mpg into mpg3, having three categories, 1 2 3, using compute and if.
compute mpg3 = 1. if (mpg >= 19) & (mpg <= 23) mpg3 = 2. if (mpg >= 24) & (mpg <= 100) mpg3 = 3. execute.
Recode mpg into mpg3a, having three categories using recode.
recode mpg (lo thru 18=1) (19 thru 23=2) (24 thru hi=3) into mpg3a.
execute.
Recode mpg into mpgfd, having two categories, but using different cutoffs for foreign and domestic cars.
compute mpgfd = $sysmis. do if foreign = 0 . recode mpg (lo thru 18=0) (19 thru hi=1) into mpgfd. end if.do if foreign = 1. recode mpg (lo thru 24=0) (25 thru hi=1) into mpgfd. end if. execute.