Calculate the local univariate join count — local_joincount

The univariate local join count statistic is used to identify clusters of rarely occurring binary variables. The binary variable of interest should occur less than half of the time.

Usage

local_joincount_uni(
  fx,
  chosen,
  listw,
  alternative = "two.sided",
  nsim = 199,
  iseed = NULL,
  no_repeat_in_row=FALSE
)

Arguments

fx: a binary variable either numeric or logical
chosen: a scalar character containing the level of fx that should be considered the observed value (1).
listw: a listw object containing binary weights created, for example, with nbwlistw(nb, style = "B")
alternative: default "greater". One of "less" or "greater".
nsim: the number of conditional permutation simulations
iseed: default NULL, used to set the seed; the output will only be reproducible if the count of CPU cores across which computation is distributed is the same
no_repeat_in_row: default FALSE, if TRUE, sample conditionally in each row without replacements to avoid duplicate values, https://github.com/r-spatial/spdep/issues/124

Value

a data.frame with two columns BB and Pr() and number of rows equal to the length of x.

Details

The local join count statistic requires a binary weights list which can be generated with nb2listw(nb, style = "B"). Additionally, ensure that the binary variable of interest is rarely occurring in no more than half of observations.

P-values are estimated using a conditional permutation approach. This creates a reference distribution from which the observed statistic is compared. For more see Geoda Glossary.

References

Anselin, L., & Li, X. (2019). Operational Local Join Count Statistics for Cluster Detection. Journal of geographical systems, 21(2), 189–210. doi:10.1007/s10109-019-00299-x

Author

Josiah Parry josiah.parry@gmail.com

Examples

data(oldcol)
fx <- as.factor(ifelse(COL.OLD$CRIME < 35, "low-crime", "high-crime"))
listw <- nb2listw(COL.nb, style = "B")
set.seed(1)
(res <- local_joincount_uni(fx, chosen = "high-crime", listw))
#>    BB Pr(z != E(BBi))
#> 1   0              NA
#> 2   0              NA
#> 3   4            0.50
#> 4   0              NA
#> 5   0              NA
#> 6   0              NA
#> 7   4            0.89
#> 8   0              NA
#> 9   5            0.11
#> 10  0              NA
#> 11  0              NA
#> 12  0              NA
#> 13  0              NA
#> 14  0              NA
#> 15  0              NA
#> 16  0              NA
#> 17  0              NA
#> 18  0              NA
#> 19  0              NA
#> 20  3            0.55
#> 21  2            0.84
#> 22  3            0.30
#> 23  4            0.61
#> 24  5            0.12
#> 25  0              NA
#> 26  0              NA
#> 27  0              NA
#> 28  0              NA
#> 29  3            0.98
#> 30  5            0.09
#> 31  8            0.01
#> 32  7            0.02
#> 33  6            0.02
#> 34  5            0.04
#> 35  7            0.03
#> 36  8            0.02
#> 37  6            0.05
#> 38  5            0.04
#> 39  3            0.44
#> 40  5            0.14
#> 41  3            0.13
#> 42  5            0.05
#> 43  2            0.63
#> 44  0              NA
#> 45  0              NA
#> 46  0              NA
#> 47  0              NA
#> 48  0              NA
#> 49  0              NA