| Title: | Compute Measures of Foreign Policy Similarity/Agreement |
| Version: | 0.1.0 |
| Maintainer: | Steven Miller <steve@svmiller.com> |
| Description: | Provides functions for calculating various measures of foreign policy similarity or association commonly used in the study of international relations. These include Signorino and Ritter's S statistic (weighted and unweighted), Cohen's weighted kappa, Scott's pi, and Kendall's tau-b. The package facilitates the generation of dyadic similarity scores for empirical analyses and can also serve as an educational resource for understanding how such measures are derived. |
| License: | GPL-2 | GPL-3 [expanded from: GPL (≥ 2)] |
| Encoding: | UTF-8 |
| RoxygenNote: | 7.3.2 |
| Depends: | R (≥ 3.6.0) |
| LazyData: | true |
| Suggests: | peacesciencer |
| NeedsCompilation: | no |
| Packaged: | 2025-07-30 10:00:01 UTC; steve |
| Author: | Steven Miller 
[aut, cre] |
| Repository: | CRAN |
| Date/Publication: | 2025-08-18 14:20:28 UTC |
Calculate Cohen's (1960, 1968) weighted kappa
Description
cohenk() takes two vectors and returns Cohen's kappa as an estimate
of chance-corrected agreement.
Usage
cohenk(x1, x2, w_exp = 2, levels = NULL)
Arguments
x1 |
a vector, and one assumes an integer |
x2 |
a vector, and one assumes an integer |
w_exp |
an exponent to apply to the weight matrix. Default is 2 for squared distances in the weight matrix. Supplying a 1 would make for linear distances. |
levels |
defaults to NULL, but an optional vector that defines the full
sequence of values that could be observed in |
Details
The function subsets to complete cases of the two vectors for which you want Cohen's kappa.
The function implicitly assumes that x1 and x2 are columns in a data
frame. One indirect check for this looks at whether x1 and x2 are the
same length. The function will stop if they're not.
There will sometimes be instances, assuredly with alliances, where not all
categories are observed. For example, the toy example I provide of Germany
and Russia in 1914 includes no 2s. In the language of "ratings", the "rating"
of 2 was available for Germany and Russia in 1914 but neither side used it.
The levels argument allows you to specify the full sequence of values that
could be observed, even if none were. It probably makes the most sense to
always use this argument, even if the default behavior operates as if you
won't.
Value
cohenk() takes two vectors and returns Cohen's kappa as an estimate
of chance-corrected agreement.
References
Cohen, Jacob. 1960. "A Coefficient of Agreement for Nominal Scales." Educational and Psychological Measurement 20(1): 37-46.
Cohen, Jacob. 1968. "Weighted Kappa: Nominal Scale Agreement with Provision for Scaled Disagreement or Partial Credit." Psychological Bulletin 70(4): 213–220.
Examples
cohenk(gmyrus14$gmy, gmyrus14$rus, levels = 0:3) # with levels argument
cohenk(usamex46$vote1, usamex46$vote2) # levels argument not necessary here.
Select Alliance Portfolios of Germany and Russia, 1914
Description
A simple example of alliance portfolios of Germany and Russia, by way of Signorino and Ritter's (1999) example.
Usage
gmyrus14
Format
A data frame with 20 observations on the following 4 variables.
statea three-character code indicating a state
syscapthe capabilities of the state, as a potential weight
gmythe alliance commitment for Germany with the state identified in the
statecolumnrusthe alliance commitment for Russia with the state identified in the
statecolumn
Details
The data come by way of Signorino and Ritter (1999, Table 6).
References
Signorino, Curtis S. and Jeffrey M. Ritter. "Tau-b or Not Tau-B: Measuring the Similarity of Foreign Policy Positions." International Studies Quarterly 43(1): 115–44.
Calculate Scott's (1955) pi
Description
spi() takes two vectors and returns Scott's (1955) pi coefficient,
communicating extent of inter-observer reliability.
Usage
spi(x1, x2, levels = NULL)
Arguments
x1 |
a vector, and one assumes an integer |
x2 |
a vector, and one assumes an integer |
levels |
defaults to NULL, but an optional vector that defines the full
sequence of values that could be observed in |
Details
The function subsets to complete cases of the two vectors for which you want Scott's pi.
The function implicitly assumes that x1 and x2 are columns in a data
frame. One indirect check for this looks at whether x1 and x2 are the
same length. The function will stop if they're not.
There will sometimes be instances, assuredly with alliances, where not all
categories are observed. For example, the toy example I provide of Germany
and Russia in 1914 includes no 2s. In the language of "ratings", the "rating"
of 2 was available for Germany and Russia in 1914 but neither side used it.
The levels argument allows you to specify the full sequence of values that
could be observed, even if none were. It probably makes the most sense to
always use this argument, even if the default behavior operates as if you
won't.
Value
spi() takes two vectors and returns Scott's (1955) pi coefficient,
communicating extent of inter-observer reliability.
References
Scott, William A. 1955. "Reliability of Content Analysis: The Case of Nominal Scale Coding." Public Opinion Quarterly 19(3): 321–5.
Examples
spi(gmyrus14$gmy, gmyrus14$rus, levels = 0:3) # with levels argument
spi(usamex46$vote1, usamex46$vote2) # levels argument not necessary here.
Calculate Signorino and Ritter's (1999) S for Similarity
Description
srs() takes two vectors and returns Signorino and Ritter's S statistic
communicating broadly understood "similarity" of interests or ratings.
Usage
srs(x1, x2, distances = "absolute", weights = NULL, range = NULL)
Arguments
x1 |
a vector, and one assumes an integer |
x2 |
a vector, and one assumes an integer |
distances |
the type of distances between ratings/attachments to estimate. Can be either "absolute" or "squared". Defaults to "absolute", but see note in details section. |
weights |
a vector of weights. Defaults to NULL for creating unweighted S statistics |
range |
defaults to NULL, but an optional vector that forces the range
to be a certain value. If NULL, the function calculates a range based on the
maximum and minimum values observed across both |
Details
Be advised that Signorino and Ritter's (1999) treatment of the S statistic used absolute distances when squared distances are more commonly used in the world of distance and association metrics.
There are potentially instances in which the conceivable range of
ratings/attachments (i.e. your two vectors) are not observed. In the case of
applications to alliance data, this is almost an impossibility. Every state,
by assumption, is maximally committed to defending itself. There will
assuredly be cases in which there is no commitment to another state in the
data (either for reasons of disinterest or enmity, though the first calls
into question what a 0 should communicate and the latter betrays the
interesting complexity of alliances). Thus, the minimum and maximum, one
assumes, will always be observed in the alliance data. Perhaps the same could
be said for UN voting data, though I couldn't rule out the possibility that
there is a dyad out there for which both states never voted "yes" or "no".
That would have implications for the range in the denominator of the formula.
You can override that by hard-setting the range in the range argument.
The function subsets to complete cases of the two vectors for which you want an S score. If weights are included, the function further subsets to complete cases including the weights as well.
The function implicitly assumes that x1 and x2 are columns in a data
frame. One indirect check for this looks at whether x1 and x2 are the
same length. The function will stop if they're not.
Several Comments on Weighting
If it were my call to make, I'd caution against the IR standard of using the composite index of national capabilities (CINC) as a weight on the calculation of the S statistic. Conceptually, weighting by capabilities tries to capture some kind of "importance" quantity. Related to the familiar application of alliances, this would prioritize those states that could conceivably bring more to the battlefield. In practice, this adds one anachronism to another. Capabilities, as measured, are basically a nineteenth century measurement for which estimates of energy consumption, iron and steel production, and urban population size are given equal weight in composition of the measure to military expenditures and military size. Alliances themselves are somewhat antiquarian, certainly in what we want them to do for this measure. If the question is "why must alliances be measures of foreign policy similarity", the answer kind of reduces to "we have historical data on them." If you want estimates for the 19th century, you have this, but then are implicitly confessing your measure of foreign policy similarity is an anachronism.
There are other peculiarities too. The data on capabilities has always been historically skewed to the right. Very few states have proportionally that much weight. As the state system has expanded in size (i.e. as empires ended), the relative weight at the top necessarily decreases. For example, the top 3 states in capabilities in 1816 (the United Kingdom, Russia, and France) combined for 61.8% of capabilities in a system of just 23 states. In 2016, the top three states (China, the United States, and India) combined for 45% of capabilities in a system of 195 states. New states are almost always small states that possess almost no capabilities. 11 of 23 states in 1816 had less than 1% of capabilities. That's about 48% of the system. In 2016, 176 of 195 states have less than 1% of capabilities. That's over 90% of the system. If the idea is to identify the "important" foreign policy ties, I echo Haege's (2011) contention that this approach is a second-best solution. It's second-best to other metrics that better model chance-corrected agreement. It just discards too much information and gives too much weight to great powers and/or states that are conspicuously high on capabilities (e.g. India).
Faithfully calculating a weighted S statistic (by system capabilities) requires a weight that sums to 1. In the most literal sense of 1, there is no year in the National Material Capabilities data (v. 2016) in which system capabilities in a given year sum to 1. In almost 60% of cases/years, the discrepancy doesn't look like a rounding error either. In 1860, all capabilities sum to over 1.07! In the context of applications with Correlates of War's CINC scores, you can still use the raw data because the function doesn't assume the weights sum to 1. You'll see how in the denominator of the formula.
Weights are only applicable to absolute distances. If you specify a weight
variable with distances = 'squared', the function will ignore your weights.
In applications to the Correlates of War system, as far as I am aware, there are no CoW states for which there isn't a CINC estimate. If, for some reason, a CINC score (or some other weight) is missing, the cases are dropped before weights are applied.
If weights are supplied, the weights must match the length of either x1 or
x2. The function builds in an implicit assumption that the weights are a
column in the data frame you're using.
Value
srs() takes two vectors and returns Signorino and Ritter's S statistic
communicating broadly understood "similarity" of interests or ratings.
References
Signorino, Curtis S. and Jeffrey M. Ritter. "Tau-b or Not Tau-B: Measuring the Similarity of Foreign Policy Positions." International Studies Quarterly 43(1): 115–44.
Examples
srs(gmyrus14$gmy, gmyrus14$rus, distances = 'absolute')
srs(gmyrus14$gmy, gmyrus14$rus, distances = 'squared')
srs(gmyrus14$gmy, gmyrus14$rus, distances = 'absolute', weights = gmyrus14$syscap)
Calculate Kendall's (1938) Tau-B
Description
taub() takes two vectors and returns Kendall's Tau-b as a measure of
rank correlation.
Usage
taub(x1, x2)
Arguments
x1 |
a vector, and one assumes an integer |
x2 |
a vector, and one assumes an integer |
Details
I'll be honest that I wrote this just to say that I did write this and that
my workflow would still lean on using the cor() function in base R.
Value
taub() takes two vectors and returns Kendall's Tau-b as a measure of
rank correlation.
References
Kendall, Maurice G. 1938. "A New Measure of Rank Correlation". Biometrika 30(1/2): 81-93.
Examples
taub(usamex46$vote1, usamex46$vote2)
taub(gmyrus14$gmy, gmyrus14$rus)
# Compare with...
cor(usamex46$vote1, usamex46$vote2, method = 'kendall')
cor(gmyrus14$gmy, gmyrus14$rus, method = 'kendall')
American-Mexican Dyadic Voting Patterns in the United Nations, 1946
Description
A simple example of voting patterns for the United States and Mexico in the United Nations in 1946.
Usage
usamex46
Format
A data frame with 38 observations on the following 6 variables.
residan identifier for a roll-call vote ID
ccode1the Correlates of War state code for the United States (2)
ccode2the Correlates of War state code for Mexico (70)
yeara numeric constant for the year (1946)
vote1an integer for how the United States voted on the resolution identified in the
residcolumnvote2an integer for how Mexico voted on the resolution identified in the
residcolumn
Details
Data are from a June 2024 of the United Nations voting data provided by Erik Voeten on his Dataverse for the project.
Valid vote values identified are 1 (yes), 2 (abstain), and 3 (no).
References
Bailey, Michael A., Anton Strezhnev, and Erik Voeten. 2017. "Estimating Dynamic State Preferences from United Nations Voting Data." Journal of Conflict Resolution 61(2): 430-56.