Computes the t-year survival probability $${}_t p_x = P[T(x) > t]$$ using an annual life table, allowing for fractional ages under standard actuarial assumptions.
Arguments
- lt
A lifetable object as produced by
lifetable. Must contain columnsxandlx. Columnsqxorpxare used if present.- x
Integer age(s) at which survival starts.
- t
Nonnegative numeric duration(s) in years (can be fractional).
- frac
Fractional-age assumption:
"UDD","CF","CML"(alias of CF), or"Balducci". If not specified andltcarries afracattribute (set bylifetable), that value is used.- tidy
Logical. If
TRUE, returns a tibble with columnsx,t,frac,tpx.- check
Logical. If
TRUE, performs validity checks.- tol
Numeric tolerance for integer checks on
x.
Details
The integer-year survival is obtained directly from the life table (Finan, Section 22): $${}_n p_x = \frac{\ell_{x+n}}{\ell_x}$$
For non-integer durations, let \(t = n + s\) with \(n = \lfloor t \rfloor\) and \(s \in [0,1)\). Then (Finan, Section 24): $${}_t p_x = {}_n p_x \times {}_s p_{x+n}$$
The fractional-year factor \({}_s p_y\) depends on the assumption:
UDD (Finan, Sec. 24.1): \({}_s p_y = 1 - s \times q_y\)
CF (Finan, Sec. 24.2): \({}_s p_y = (p_y)^s\)
Balducci (Finan, Sec. 24.3): \({}_s p_y = \frac{p_y}{1 - (1 - s) \times q_y}\)
Missing annual survival probabilities are derived directly from adjacent
survivor counts, avoiding cancellation in \(1-q_y\). CF and Balducci
retain logarithmic survival when that count ratio underflows. Supplied
px values remain in use; input columns should describe a consistent
annual life table.
If \(x + t > \omega\) (the terminal age), the function returns 0
by the established table-horizon convention, including open tables.
This is truncation, not inference about survival beyond observed ages;
extend the table before evaluating a longer horizon. Interval tables with
nonannual widths from km_lifetable require annual conversion first.