README.md

In danielwilhelm/R-CS-ranks: Statistical Tools for Ranks

csranks

The R package csranks provides statistical tools when working with ranks. Two central functions are csranks for confidence sets and lmranks for linear regression.

csranks implements confidence sets for positions of populations in a ranking based on values of a certain feature and their estimation errors. Both simultaneous and marginal confidence sets are available, as well as confidence sets with populations occupying top-n positions in the ranking. Work based on Mogstad, Romano, Shaikh, and Wilhelm (2023).

lmranks implements linear regression with consistent estimation of standard errors and covariance matrix of regression parameters $\beta$.

Installation

You can install the released version of csranks from CRAN with:

install.packages("csranks")

You can install the development version of csranks from GitHub with:

# install.packages("remotes")
remotes::install_github("danielwilhelm/R-CS-ranks")

Example: PISA Ranking

This example illustrates the computation of confidence sets for the ranking of countries according students’ achievements in mathematics. The following dataset contains data from the 2018 Program for International Student Assessment (PISA) study by the Organization for Economic Cooperation and Development (OECD), which can be downloaded at https://www.oecd.org/pisa/data/. Students’ achievement is measured by the average mathematics test scores for all 15-year-old students in the study.

First, load the package csranks and the data:

library(csranks)
attach(pisa)

The dataframe pisa looks like this:

head(pisa)
#>   jurisdiction science_score science_se reading_score reading_se math_score
#> 1    Australia      502.9646   1.795398      502.6317   1.634343   491.3600
#> 2      Austria      489.7804   2.777395      484.3926   2.697472   498.9423
#> 3      Belgium      498.7731   2.229240      492.8644   2.321973   508.0703
#> 4       Canada      517.9977   2.153651      520.0855   1.799716   512.0169
#> 5        Chile      443.5826   2.415280      452.2726   2.643766   417.4066
#> 6     Colombia      413.3230   3.052402      412.2951   3.251344   390.9323
#>    math_se
#> 1 1.939833
#> 2 2.970999
#> 3 2.262662
#> 4 2.357476
#> 5 2.415888
#> 6 2.989559

Consider ranking countries according to their students’ achievement in mathematics. The scores in math_score are estimates of these achievements. The countries’ ranks can be estimated using the function irank():

math_rank <- irank(math_score)
math_rank
#>  [1] 24 18 10  7 35 37 17  8  3 11 20 15 34 30 21 16 32 25  1  2 19 29 27 36  4
#> [26] 22 14  5 23 26  9 28 12  6 33 13 31

These are only estimated ranks as they are computed from estimates of achievements and therefore estimation uncertainty in the estimates translates to estimation uncertainty in the ranks. The following subsections discuss two different confidence sets for assessing such estimation uncertainty in ranks: marginal and simultaneous confidence sets. Explanations and details about the differences of these two confidence sets can be found in Mogstad, Romano, Shaikh, and Wilhelm (2020). The complete R script containing this example, as well as R scripts for PISA rankings by reading and science scores, can be found in the subdirectory examples/.

Marginal Confidence Sets

The marginal confidence sets for the ranks are computed as follows:

math_cov_mat <- diag(math_se^2)
CS_marg <- csranks(math_score, math_cov_mat, coverage=0.95, simul=FALSE, R=1000, seed=101)
math_rankL_marg <- CS_marg$L
math_rankU_marg <- CS_marg$U

where math_rankL_marg and math_rankL_marg contain the lower and upper bounds of the confidence sets for the ranks. They can be plotted by

grid::current.viewport()
#> viewport[ROOT]

plotmarg <- plot(CS_marg, popnames=jurisdiction, title="Ranking of OECD Countries by 2018 PISA Math Score", 
    subtitle="(with 95% marginal confidence sets)", colorbins=4)
plotmarg

You can then save the graph by

ggplot2::ggsave("mathmarg.pdf", plot=plotmarg)

Simultaneous Confidence Sets

The simultaneous confidence sets for the ranks are computed as follows:

CS_simul <- csranks(math_score, math_cov_mat, coverage=0.95, simul=TRUE, R=1000, seed=101)
math_rankL_simul <- CS_simul$L
math_rankU_simul <- CS_simul$U

where math_rankL_simul and math_rankL_simul contain the lower and upper bounds of the confidence sets for the ranks. They can be plotted by

grid::current.viewport()
#> viewport[ROOT]

plotsimul <- plot(CS_simul, popnames=jurisdiction, title="Ranking of OECD Countries by 2018 PISA Math Score", 
    subtitle="(with 95% simultaneous confidence sets)", colorbins=4)

plotsimul

You can then save the graph by

ggplot2::ggsave("mathsimul.pdf", plot=plotsimul)

Reference

Mogstad, Romano, Shaikh, and Wilhelm (2023), “Inference for Ranks with Applications to Mobility across Neighborhoods and Academic Achievements across Countries”, forthcoming at Review of Economic Studies

danielwilhelm/R-CS-ranks documentation built on Sept. 18, 2023, 11:38 p.m.