Formula: P(new) = (alpha + sigma K) / (n + alpha)
Usage
Dpparit(n = 100, alpha = 1, sigma = 0.5, seed = 42)
Arguments
- n
Number of customers to seat.
- alpha
Concentration; must exceed -sigma.
- sigma
Discount in [0, 1).
- seed
Seed of the deterministic stream.
Value
List with estimate (number of blocks), K,
counts, p_new, n, method.
Details
An occupied table takes (n_k - sigma)/(n + alpha), so the discount
sigma moves mass from large tables to new ones and the number of
blocks grows like n^sigma instead of log n. Setting sigma = 0
recovers the Dirichlet process exactly.
References
Pitman & Yor (1997), Ann. Probab. 25(2):855-900.
Examples
Dpparit()
#> $estimate
#> [1] 10
#>
#> $K
#> [1] 10
#>
#> $counts
#> [1] 58 26 1 7 2 1 1 2 1 1
#>
#> $p_new
#> [1] 0.05940594
#>
#> $n
#> [1] 100
#>
#> $method
#> [1] "Pitman-Yor two-parameter seating process"
#>
#> attr(,"class")
#> [1] "morie_rich_result" "list"
