In this chapter, we want to continue working on the challenge introduced in Chapter 12 of calculating Theil’s H - a measure of racial segregation - for every county in the United States. If you haven’t read that chapter, you should probably do so before continuing on here.
The dataset we used for this analysis is based on tract level data of population counts by race extracted from Social Explorer. You can see a snippet below:
In Chapter 12, we developed a handy function for calculating Theil’s H from census tract data for a given county. This saves us a lot of time and code, because that function allows us to reuse the code for calculating Theil’s H for any county we want. For example, if I want the Theil’s H of Alameda County, California:
tracts |>filter(county_name =="Alameda County, California") |>calc_theil_h()
[1] 0.1689115
However, we still have a bit of a problem. If I want to calculate Theil’s H for all 3,222 counties in the dataset, entering them in one by one is going to be extremely laborious. How can I get R to do the hard work here?
Computers are great at iteration. We basically want to tell R to repeat the same procedure over and over again across all counties in my dataset. There are two approaches to iteration in R: loops and mapping.
Looping
Looping is a common procedure in most computer programming languages. In this case, we want a for loop which will iterate across a vector of values and repeat the same code.1 Here is a simple example of a for loop:
The value i is a placeholder that can be referenced within the for loop itself. The first time the for loop is run, it uses the value of 1 and processes the code in the curly brackets. The next time, the for loop is run, it uses the value of 2, and so on until the final value of 10 is reached.
We want to loop across every county. In order to do that, I need a unique identifier for a county. I could use name or ID in this case. Generally, names are not as reliable, but for more intuitive display, I will use names here as I know that no county names are duplicated within states. To get the unique county names, I can use the unique function:
For expository purposes, I am going to take a small sample of these counties to iterate across so that you don’t have to look at output for all 3,222 counties. Once we get the code working for the small sample, we will go back to the full set.
counties_sample <-sample(counties, 10)
First, lets test out a simple for loop where we print out county names:
for(county in counties_sample) {print(county)}
[1] "Alger County, Michigan"
[1] "Kewaunee County, Wisconsin"
[1] "Catahoula Parish, Louisiana"
[1] "Marshall County, Mississippi"
[1] "Jackson County, Missouri"
[1] "Dakota County, Nebraska"
[1] "Jones County, North Carolina"
[1] "Prince Edward County, Virginia"
[1] "Tipton County, Tennessee"
[1] "Orange County, New York"
Note that I used county rather than i as my placeholder value name because its more intuitive. You can use any name you like here. The for loop seems to be working, so lets try to actually calculate Theil’s H within the loop.
for(county in counties_sample) { tracts |>filter(county_name == county) |>calc_theil_h() |>print()}
That seemed to work. However, at this point, I am just spitting out the results to the screen. I would prefer to save these results back to something. I could just save the results back to a vector, but I would rather save them back to a tibble that includes both the county name and Theil’s H. Either way, I do this by initializing a NULL object and then adding to that object within the loop. For a vector, I can do this by simply concatenating the new value onto the old with c:
theil_h <-NULLfor(county in counties_sample) { h <- tracts |>filter(county_name == county) |>calc_theil_h() theil_h <-c(theil_h, h)}theil_h
However, when I only return a vector, I lose information on which value belongs to which county. Instead, I am going to save the results to a tibble and then use the bind_rows command to add this tibble to my existing tibble:
theil_h <-NULLfor(county in counties_sample) {# calculate Theil's H for this county h <- tracts |>filter(county_name == county) |>calc_theil_h()# add this county's values to our dataset of values theil_h <- theil_h |>bind_rows(tibble(county_name = county, theil_h = h))}theil_h
# A tibble: 10 × 2
county_name theil_h
<chr> <dbl>
1 Alger County, Michigan 0.136
2 Kewaunee County, Wisconsin 0.0663
3 Catahoula Parish, Louisiana 0.0982
4 Marshall County, Mississippi 0.143
5 Jackson County, Missouri 0.228
6 Dakota County, Nebraska 0.136
7 Jones County, North Carolina 0.0213
8 Prince Edward County, Virginia 0.0493
9 Tipton County, Tennessee 0.129
10 Orange County, New York 0.145
One thing to keep in mind is that if you end up rerunning your for loop, you also need to remember to re-initialize your object of returned value as a NULL object. Otherwise, your results will be added on to the pre-existing object.
Now that I have this code working for a small sample, lets run it for all 3,222 counties.
theil_h <-NULLfor(county in counties) {# calculate Theil's H for this county h <- tracts |>filter(county_name == county) |>calc_theil_h()# add this county's values to our dataset of values theil_h <- theil_h |>bind_rows(tibble(county_name = county, theil_h = h))}theil_h
# A tibble: 3,222 × 2
county_name theil_h
<chr> <dbl>
1 Autauga County, Alabama 0.120
2 Baldwin County, Alabama 0.117
3 Barbour County, Alabama 0.0625
4 Bibb County, Alabama 0.155
5 Blount County, Alabama 0.129
6 Bullock County, Alabama 0.110
7 Butler County, Alabama 0.116
8 Calhoun County, Alabama 0.188
9 Chambers County, Alabama 0.107
10 Cherokee County, Alabama 0.116
# ℹ 3,212 more rows
This is too many values to display, so lets go ahead and look at a histogram of Theil’s H across counties in the US:
ggplot(theil_h, aes(x = theil_h))+geom_histogram(color ="grey20", fill ="grey70")+theme_bw()+labs(x ="Theil's H")
Figure 13.1: Distribution of Theil’s H across US counties.
Mapping
For loops are fairly intuitive but have a bad reputation in R. Historically, for loops in R have been much slower than in many other programming languages (for reasons are that are technically complex). For that reason, many people prefer to use the methods of mapping rather than looping in R for speedier results. In more recent version of R, looping has improved considerably and so the advantages of mapping may be somewhat overstated. Nonetheless, it is useful to know both methods as mapping can sometimes offer other efficiencies over looping, or provide more compact code.
Our primary tool is going to be the map function from the purrr package. However, you can also use the lapply function from base R in the same manner. Both functions allow you to apply some other function to every element of a list.
Whats a list, you say? You may want to Chapter 3, but the short answer is that lists are collections of objects. The objects within lists can be of different types, but for the purposes of map and lapply we want lists that contain objects of the same type, since we are going to apply the same function to each element of the list.
The first step to getting map to work is to create the list. In our case, we want to split up our single dataset of all tracts to a list of datasets where each county is a separate element of the list. Each dataset within the list will have the same structure.
To split a dataset into a list of datasets, we can use the group_by command to define what variable should be used to split the list (county name and id, in this case), followed by a group_split command. For expository purposes, rather than splitting the full tract dataset, I am going to use the sample of ten counties from above.
The first element is the dataset for Alger County, Michigan. Each of the ten elements contains a dataset for a given county.
Now that I have my data structured the way that map wants the input, I can use the map function itself. The other required argument for map is a function that will be applied to each element of the list. I can use a pre-defined function or I can create a custom function. I will first demonstrate how it works by getting a summary for each dataset:
map(county_list, summary)
[[1]]
tract_id county_id county_name pop_total
Min. :2.6e+10 Min. :26003 Length:3 Min. :2203
1st Qu.:2.6e+10 1st Qu.:26003 Class :character 1st Qu.:2696
Median :2.6e+10 Median :26003 Mode :character Median :3190
Mean :2.6e+10 Mean :26003 Mean :2955
3rd Qu.:2.6e+10 3rd Qu.:26003 3rd Qu.:3332
Max. :2.6e+10 Max. :26003 Max. :3473
pop_race_white pop_race_black pop_race_asian pop_race_other pop_race_multi
Min. :1925 Min. : 2.0 Min. : 2.00 Min. :0 Min. : 79
1st Qu.:2126 1st Qu.: 10.0 1st Qu.: 2.00 1st Qu.:0 1st Qu.: 91
Median :2328 Median : 18.0 Median : 2.00 Median :0 Median :103
Mean :2422 Mean :230.3 Mean :23.33 Mean :2 Mean :125
3rd Qu.:2670 3rd Qu.:344.5 3rd Qu.:34.00 3rd Qu.:3 3rd Qu.:148
Max. :3013 Max. :671.0 Max. :66.00 Max. :6 Max. :193
pop_race_latino pop_race_indigenous
Min. : 28.00 Min. : 18.00
1st Qu.: 34.50 1st Qu.: 30.50
Median : 41.00 Median : 43.00
Mean : 58.33 Mean : 94.33
3rd Qu.: 73.50 3rd Qu.:132.50
Max. :106.00 Max. :222.00
[[2]]
tract_id county_id county_name pop_total
Min. :2.203e+10 Min. :22025 Length:3 Min. :1575
1st Qu.:2.203e+10 1st Qu.:22025 Class :character 1st Qu.:2122
Median :2.203e+10 Median :22025 Mode :character Median :2670
Mean :2.203e+10 Mean :22025 Mean :2965
3rd Qu.:2.203e+10 3rd Qu.:22025 3rd Qu.:3660
Max. :2.203e+10 Max. :22025 Max. :4650
pop_race_white pop_race_black pop_race_asian pop_race_other pop_race_multi
Min. : 509 Min. : 448.0 Min. :0.000 Min. :0 Min. : 0.00
1st Qu.:1362 1st Qu.: 739.0 1st Qu.:0.000 1st Qu.:0 1st Qu.:14.50
Median :2215 Median :1030.0 Median :0.000 Median :0 Median :29.00
Mean :1962 Mean : 935.3 Mean :2.333 Mean :0 Mean :29.67
3rd Qu.:2688 3rd Qu.:1179.0 3rd Qu.:3.500 3rd Qu.:0 3rd Qu.:44.50
Max. :3162 Max. :1328.0 Max. :7.000 Max. :0 Max. :60.00
pop_race_latino pop_race_indigenous
Min. : 0.00 Min. : 0.000
1st Qu.: 3.50 1st Qu.: 0.000
Median : 7.00 Median : 0.000
Mean :32.33 Mean : 3.333
3rd Qu.:48.50 3rd Qu.: 5.000
Max. :90.00 Max. :10.000
[[3]]
tract_id county_id county_name pop_total
Min. :3.104e+10 Min. :31043 Length:5 Min. :2899
1st Qu.:3.104e+10 1st Qu.:31043 Class :character 1st Qu.:3117
Median :3.104e+10 Median :31043 Mode :character Median :3928
Mean :3.104e+10 Mean :31043 Mean :4262
3rd Qu.:3.104e+10 3rd Qu.:31043 3rd Qu.:4535
Max. :3.104e+10 Max. :31043 Max. :6829
pop_race_white pop_race_black pop_race_asian pop_race_other pop_race_multi
Min. : 879 Min. : 9.0 Min. : 63 Min. :0 Min. : 52.0
1st Qu.:1037 1st Qu.: 15.0 1st Qu.: 63 1st Qu.:0 1st Qu.: 76.0
Median :2430 Median :160.0 Median : 79 Median :0 Median : 85.0
Mean :1919 Mean :299.2 Mean :136 Mean :0 Mean :119.4
3rd Qu.:2514 3rd Qu.:593.0 3rd Qu.:214 3rd Qu.:0 3rd Qu.:148.0
Max. :2737 Max. :719.0 Max. :261 Max. :0 Max. :236.0
pop_race_latino pop_race_indigenous
Min. : 151 Min. : 0.0
1st Qu.: 970 1st Qu.: 24.0
Median :1710 Median : 40.0
Mean :1696 Mean : 91.4
3rd Qu.:1872 3rd Qu.: 83.0
Max. :3778 Max. :310.0
[[4]]
tract_id county_id county_name pop_total
Min. :2.91e+10 Min. :29095 Length:224 Min. : 23
1st Qu.:2.91e+10 1st Qu.:29095 Class :character 1st Qu.:2051
Median :2.91e+10 Median :29095 Mode :character Median :3042
Mean :2.91e+10 Mean :29095 Mean :3194
3rd Qu.:2.91e+10 3rd Qu.:29095 3rd Qu.:3948
Max. :2.91e+10 Max. :29095 Max. :7356
pop_race_white pop_race_black pop_race_asian pop_race_other
Min. : 6.0 Min. : 0.0 Min. : 0.00 Min. : 0.00
1st Qu.: 866.2 1st Qu.: 169.8 1st Qu.: 0.00 1st Qu.: 0.00
Median :1702.5 Median : 507.5 Median : 29.00 Median : 0.00
Mean :1952.6 Mean : 723.0 Mean : 55.53 Mean : 16.65
3rd Qu.:2757.2 3rd Qu.:1051.5 3rd Qu.: 82.25 3rd Qu.: 11.00
Max. :5710.0 Max. :3364.0 Max. :421.00 Max. :316.00
pop_race_multi pop_race_latino pop_race_indigenous
Min. : 0.00 Min. : 0.0 Min. : 0.00
1st Qu.: 58.75 1st Qu.: 90.5 1st Qu.: 0.00
Median :103.00 Median : 183.5 Median : 0.00
Mean :126.01 Mean : 306.1 Mean : 14.39
3rd Qu.:166.00 3rd Qu.: 346.2 3rd Qu.: 13.00
Max. :683.00 Max. :2737.0 Max. :403.00
[[5]]
tract_id county_id county_name pop_total
Min. :3.71e+10 Min. :37103 Length:3 Min. :2776
1st Qu.:3.71e+10 1st Qu.:37103 Class :character 1st Qu.:2930
Median :3.71e+10 Median :37103 Mode :character Median :3083
Mean :3.71e+10 Mean :37103 Mean :3088
3rd Qu.:3.71e+10 3rd Qu.:37103 3rd Qu.:3244
Max. :3.71e+10 Max. :37103 Max. :3404
pop_race_white pop_race_black pop_race_asian pop_race_other
Min. :1727 Min. : 642.0 Min. : 4.00 Min. :0
1st Qu.:1854 1st Qu.: 712.0 1st Qu.:20.50 1st Qu.:0
Median :1981 Median : 782.0 Median :37.00 Median :0
Mean :1913 Mean : 854.7 Mean :31.67 Mean :0
3rd Qu.:2006 3rd Qu.: 961.0 3rd Qu.:45.50 3rd Qu.:0
Max. :2031 Max. :1140.0 Max. :54.00 Max. :0
pop_race_multi pop_race_latino pop_race_indigenous
Min. : 19.00 Min. : 86.0 Min. : 9.00
1st Qu.: 53.50 1st Qu.:144.0 1st Qu.:18.00
Median : 88.00 Median :202.0 Median :27.00
Mean : 90.33 Mean :176.3 Mean :21.67
3rd Qu.:126.00 3rd Qu.:221.5 3rd Qu.:28.00
Max. :164.00 Max. :241.0 Max. :29.00
[[6]]
tract_id county_id county_name pop_total
Min. :5.506e+10 Min. :55061 Length:5 Min. :2370
1st Qu.:5.506e+10 1st Qu.:55061 Class :character 1st Qu.:3260
Median :5.506e+10 Median :55061 Mode :character Median :3724
Mean :5.506e+10 Mean :55061 Mean :4114
3rd Qu.:5.506e+10 3rd Qu.:55061 3rd Qu.:5212
Max. :5.506e+10 Max. :55061 Max. :6004
pop_race_white pop_race_black pop_race_asian pop_race_other pop_race_multi
Min. :2208 Min. : 0.0 Min. : 0 Min. : 0 Min. : 37.0
1st Qu.:2959 1st Qu.: 0.0 1st Qu.: 6 1st Qu.: 0 1st Qu.: 48.0
Median :3522 Median : 3.0 Median : 6 Median : 2 Median : 68.0
Mean :3848 Mean :19.4 Mean :15 Mean : 9 Mean : 75.2
3rd Qu.:4725 3rd Qu.:16.0 3rd Qu.:20 3rd Qu.: 4 3rd Qu.: 85.0
Max. :5827 Max. :78.0 Max. :43 Max. :39 Max. :138.0
pop_race_latino pop_race_indigenous
Min. : 46 Min. : 0.0
1st Qu.: 81 1st Qu.: 0.0
Median :100 Median : 0.0
Mean :139 Mean : 8.2
3rd Qu.:143 3rd Qu.: 5.0
Max. :325 Max. :36.0
[[7]]
tract_id county_id county_name pop_total
Min. :2.809e+10 Min. :28093 Length:10 Min. :2164
1st Qu.:2.809e+10 1st Qu.:28093 Class :character 1st Qu.:2742
Median :2.809e+10 Median :28093 Mode :character Median :3594
Mean :2.809e+10 Mean :28093 Mean :3398
3rd Qu.:2.809e+10 3rd Qu.:28093 3rd Qu.:3860
Max. :2.809e+10 Max. :28093 Max. :4617
pop_race_white pop_race_black pop_race_asian pop_race_other pop_race_multi
Min. : 375.0 Min. : 558 Min. : 0 Min. : 0.0 Min. : 0.00
1st Qu.: 998.2 1st Qu.: 918 1st Qu.: 0 1st Qu.: 0.0 1st Qu.: 27.75
Median :1613.5 Median :1258 Median : 0 Median : 0.0 Median : 91.50
Mean :1600.6 Mean :1548 Mean : 4 Mean :16.7 Mean : 82.60
3rd Qu.:2111.8 3rd Qu.:1989 3rd Qu.: 0 3rd Qu.:21.0 3rd Qu.:113.75
Max. :3098.0 Max. :3470 Max. :30 Max. :99.0 Max. :165.00
pop_race_latino pop_race_indigenous
Min. : 0.00 Min. : 0.0
1st Qu.: 27.75 1st Qu.: 0.0
Median :112.50 Median : 0.0
Mean :145.20 Mean : 1.4
3rd Qu.:183.00 3rd Qu.: 0.0
Max. :477.00 Max. :11.0
[[8]]
tract_id county_id county_name pop_total
Min. :3.607e+10 Min. :36071 Length:92 Min. : 1568
1st Qu.:3.607e+10 1st Qu.:36071 Class :character 1st Qu.: 3327
Median :3.607e+10 Median :36071 Mode :character Median : 4348
Mean :3.607e+10 Mean :36071 Mean : 4361
3rd Qu.:3.607e+10 3rd Qu.:36071 3rd Qu.: 5130
Max. :3.607e+10 Max. :36071 Max. :10758
pop_race_white pop_race_black pop_race_asian pop_race_other
Min. : 246 Min. : 0.0 Min. : 0.00 Min. : 0.00
1st Qu.: 1858 1st Qu.: 118.0 1st Qu.: 24.25 1st Qu.: 0.00
Median : 2556 Median : 371.5 Median : 60.50 Median : 5.50
Mean : 2637 Mean : 455.9 Mean :123.46 Mean : 21.59
3rd Qu.: 3354 3rd Qu.: 623.5 3rd Qu.:180.50 3rd Qu.: 32.50
Max. :10080 Max. :2405.0 Max. :591.00 Max. :256.00
pop_race_multi pop_race_latino pop_race_indigenous
Min. : 0.0 Min. : 0.0 Min. : 0.000
1st Qu.: 60.0 1st Qu.: 524.0 1st Qu.: 0.000
Median :121.0 Median : 892.5 Median : 0.000
Mean :144.5 Mean : 972.8 Mean : 5.663
3rd Qu.:186.8 3rd Qu.:1320.0 3rd Qu.: 0.000
Max. :618.0 Max. :3221.0 Max. :76.000
[[9]]
tract_id county_id county_name pop_total
Min. :5.115e+10 Min. :51147 Length:6 Min. :3124
1st Qu.:5.115e+10 1st Qu.:51147 Class :character 1st Qu.:3178
Median :5.115e+10 Median :51147 Mode :character Median :3394
Mean :5.115e+10 Mean :51147 Mean :3654
3rd Qu.:5.115e+10 3rd Qu.:51147 3rd Qu.:3488
Max. :5.115e+10 Max. :51147 Max. :5385
pop_race_white pop_race_black pop_race_asian pop_race_other
Min. :1636 Min. : 702 Min. : 0.00 Min. : 0.000
1st Qu.:1770 1st Qu.: 827 1st Qu.: 27.25 1st Qu.: 0.000
Median :1966 Median :1152 Median : 29.50 Median : 0.000
Mean :2212 Mean :1087 Mean : 51.33 Mean : 1.833
3rd Qu.:2394 3rd Qu.:1286 3rd Qu.: 68.50 3rd Qu.: 0.000
Max. :3461 Max. :1466 Max. :141.00 Max. :11.000
pop_race_multi pop_race_latino pop_race_indigenous
Min. : 0.00 Min. : 46.0 Min. : 0.00
1st Qu.: 65.25 1st Qu.: 60.5 1st Qu.: 7.50
Median :113.00 Median : 97.0 Median : 23.50
Mean :125.83 Mean :131.8 Mean : 44.33
3rd Qu.:157.00 3rd Qu.:125.2 3rd Qu.: 29.75
Max. :308.00 Max. :368.0 Max. :185.00
[[10]]
tract_id county_id county_name pop_total
Min. :4.717e+10 Min. :47167 Length:13 Min. : 1566
1st Qu.:4.717e+10 1st Qu.:47167 Class :character 1st Qu.: 2783
Median :4.717e+10 Median :47167 Mode :character Median : 4849
Mean :4.717e+10 Mean :47167 Mean : 4701
3rd Qu.:4.717e+10 3rd Qu.:47167 3rd Qu.: 5602
Max. :4.717e+10 Max. :47167 Max. :10769
pop_race_white pop_race_black pop_race_asian pop_race_other
Min. :1114 Min. : 55.0 Min. : 0.00 Min. : 0.00
1st Qu.:2028 1st Qu.: 286.0 1st Qu.: 0.00 1st Qu.: 0.00
Median :2983 Median : 520.0 Median : 4.00 Median : 0.00
Mean :3504 Mean : 869.5 Mean : 26.23 Mean : 31.31
3rd Qu.:4603 3rd Qu.:1237.0 3rd Qu.: 26.00 3rd Qu.: 16.00
Max. :8682 Max. :2938.0 Max. :120.00 Max. :169.00
pop_race_multi pop_race_latino pop_race_indigenous
Min. : 32.0 Min. : 0.0 Min. : 0.000
1st Qu.: 70.0 1st Qu.: 35.0 1st Qu.: 0.000
Median : 94.0 Median : 96.0 Median : 0.000
Mean :127.7 Mean :138.2 Mean : 4.154
3rd Qu.:150.0 3rd Qu.:161.0 3rd Qu.: 2.000
Max. :445.0 Max. :454.0 Max. :26.000
This is a lot of information! By default, map returns a list of the same dimension as the input (in this case, 10 elements). In some cases, we can simplify this output, but not in this case.
Lets try creating a custom function within map itself. In this case, I want to get the sum of pop_total within each dataset.
In this case, I am actually writing the code for a custom function within another function. In this case, my function is quite simple, but you can use this feature to do some quite complex stuff if you want.
Again, the output is formatted as a list. However, I am returning a simple number for each value, so I would prefer that this output just be displayed as a simple vector of results. You can do this by using a variety of map_* functions where the second part indicates the return type you expect. In this case, I can use map_dbl because I expect the output to be a numeric value.2 Lets try it:
That seemed to work perfectly and gives me the same output as the for loop above. However, I would also like to get the county information returned in a table format. How, can I do that? I can use a custom function to output a tibble that includes the county name and id in the result:
[[1]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Alger County, Michigan 26003 0.136
[[2]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Catahoula Parish, Louisiana 22025 0.0982
[[3]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Dakota County, Nebraska 31043 0.136
[[4]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Jackson County, Missouri 29095 0.228
[[5]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Jones County, North Carolina 37103 0.0213
[[6]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Kewaunee County, Wisconsin 55061 0.0663
[[7]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Marshall County, Mississippi 28093 0.143
[[8]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Orange County, New York 36071 0.145
[[9]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Prince Edward County, Virginia 51147 0.0493
[[10]]
# A tibble: 1 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Tipton County, Tennessee 47167 0.129
This approach worked, but because I am returning a tibble each time, I am back to the list format for my output. What I want is each of these results as rows of a shared tibble. Luckily, bind_rows will do this for me. I just have to pipe the results of map into a bind_rows to get everything formatted nicely:
# A tibble: 10 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Alger County, Michigan 26003 0.136
2 Catahoula Parish, Louisiana 22025 0.0982
3 Dakota County, Nebraska 31043 0.136
4 Jackson County, Missouri 29095 0.228
5 Jones County, North Carolina 37103 0.0213
6 Kewaunee County, Wisconsin 55061 0.0663
7 Marshall County, Mississippi 28093 0.143
8 Orange County, New York 36071 0.145
9 Prince Edward County, Virginia 51147 0.0493
10 Tipton County, Tennessee 47167 0.129
Now the format is working perfectly. Lets go ahead and apply this to the full dataset in one pipe:
# A tibble: 3,222 × 3
county_name county_id theil_h
<chr> <dbl> <dbl>
1 Abbeville County, South Carolina 45001 0.127
2 Acadia Parish, Louisiana 22001 0.263
3 Accomack County, Virginia 51001 0.129
4 Ada County, Idaho 16001 0.110
5 Adair County, Iowa 19001 0.0902
6 Adair County, Kentucky 21001 0.0768
7 Adair County, Missouri 29001 0.0843
8 Adair County, Oklahoma 40001 0.0424
9 Adams County, Colorado 8001 0.122
10 Adams County, Idaho 16003 0.0240
# ℹ 3,212 more rows
Students often go with the for loop approach because it feels more natural and intuitive. In particular, students are often scared off by the custom function business. However, mapping offers you more flexibility and often speed, and so its worth learning. In this case, it was much easier to get both county name and id in the final output using mapping rather than looping.
We can also wrap each command in system.time to see how they actually perform in terms of time.
theil_h <-NULLsystem.time(for(county in counties) {# calculate Theil's H for this county h <- tracts |>filter(county_name == county) |>calc_theil_h()# add this county's values to our dataset of values theil_h <- theil_h |>bind_rows(tibble(county_name = county, theil_h = h)) })
The difference in speed is pretty minimal here. The for loop takes about 10% more time, which amounts to 2 seconds in this case. For many operations, this gain is probably not enough to warrant a decision on looping vs. mapping. However, for computationally intensive operations, a 10% savings in time may make a big difference in absolute time.
The other option is a while loop which will repeat the same code until a condition is met.↩︎
The equivalent approach to lapply is to use sapply which simplifies the list to a vector if possible.↩︎
