Interpolation, makes it happen.

lumar.R

library(zoo)

path <- "~/Desktop/Total Deaths - Assault - All Ages _ tom.csv"
df <- read.csv(path, na.strings = "...", skip = 1)
# transpose the dataframe (flip it)
df <- setNames(data.frame(t(df[,-1])), df[,1])
# This just makes the names prettier
names(df) <- gsub("^ ", "", gsub("\\d", "", names(df)))
# Falkland Islands only has one value so there's nothing to interpolate
# so we'll set them all to 0
df$`Falkland Islands (Malvinas)` <- 0
# This is where the magic happens
df_i <- data.frame( # create a data_frame called df_i
  lapply(df, # to each column apply the following function
         # This function approximates the values, and then rounds it to the nearest
         # whole number
         function(x) round(na.approx(x, rule = 2))))

# Make the row and column names pretty again
names(df_i) <- names(df)
row.names(df_i) <- 1983:2012

# write out the results
write.csv(df_i, file = "assaults_interpolated.csv")