Interpolate data using indicator in time series

You can use a linear regression to predict the missing values:

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
df  
  pivotlonger(-year) %>%
  ggplot(aes(x = year, y = value, color = name)) +
  geomline() +
  geompoint() +
  # highlight original values
  geompoint(data = ~filter(.x, name == "Variable"),
             size = 3) +
  labs(x = "Year", y = "Value", color = "Variable") +
  guides(color = guidelegend(reverse = T))
#> Warning: Removed 2 rows containing missing values or values outside the scale range
#> (geomline()).
#> Warning: Removed 8 rows containing missing values or values outside the scale range
#> (geompoint()).
#> Removed 8 rows containing missing values or values outside the scale range
#> (geompoint()).

^{Created on 2025-11-03 with reprex v2.1.1}

The na.spline function in the zoo package will take a zoo object and replace the NAs using a cubic spline interpolation.

We could run df$Varspline = zoo::na.spline(zoo::zoo(df$Variable, df$year)) and then see this works nicely with the data, interpolating between the known points and extrapolating beyond:

ggplot(df, aes(year, Varspline)) + 
  geomline(color = "gray70") +
  geompoint(shape = 21) +
  geom_point(aes(y = Variable))