在指定条件下计算缺失观测数的平均值

Question

I need to calculate the mean value for each country from 1991-2000, but only if that country is missing 2 or less years (NA) of data in that time range.

So, here's a sample of the data I have...

# A tibble: 275 x 52
   country    aid1960 aid1961 aid1962  aid1963 aid1964 aid1965 aid1966 aid1967  aid1968
   <chr>        <dbl>   <dbl>   <dbl>    <dbl>   <dbl>   <dbl>   <dbl>   <dbl>    <dbl>
 1 Abkhazia   NA        NA      NA    NA         NA     NA      NA       NA    NA      
 2 Afghanist~  1.78      3.52    1.68  3.57       4.41   5.04    4.61     3.64  2.56   
 3 Akrotiri ~ NA        NA      NA    NA         NA     NA      NA       NA    NA      
 4 Albania    NA        NA      NA    NA         NA     NA      NA       NA    NA      
 5 Algeria    32.9      39.8    35.5  24.7       19.5   12.3    10.4      8.56  9.24   
 6 American ~ NA        NA      NA    NA         NA     NA      NA       NA    NA      
 7 Andorra    NA        NA      NA    NA         NA     NA      NA       NA    NA      
 8 Angola     -0.0101    4.66   NA     0.00572   NA      0.204   0.518    3.24  0.00175
 9 Anguilla   NA        NA      NA    NA         NA     NA      NA       NA    NA      
10 Antigua a~ NA        NA      NA    NA         NA     NA      NA       NA    NA      
# ... with 265 more rows, and 42 more variables: aid1969 <dbl>, aid1970 <dbl>,
#   aid1971 <dbl>, aid1972 <dbl>, aid1973 <dbl>, aid1974 <dbl>, aid1975 <dbl>,
#   aid1976 <dbl>, aid1977 <dbl>, aid1978 <dbl>, aid1979 <dbl>, aid1980 <dbl>,
#   aid1981 <dbl>, aid1982 <dbl>, aid1983 <dbl>, aid1984 <dbl>, aid1985 <dbl>,
#   aid1986 <dbl>, aid1987 <dbl>, aid1988 <dbl>, aid1989 <dbl>, aid1990 <dbl>,
#   aid1991 <dbl>, aid1992 <dbl>, aid1993 <dbl>, aid1994 <dbl>, aid1995 <dbl>,
#   aid1996 <dbl>, aid1997 <dbl>, aid1998 <dbl>, aid1999 <dbl>, aid2000 <dbl>, ...

I was able to determine which countries were missing data in that time range using this code...

rowSums(is.na(countrydata[, 33:42]))

Which gave me an output of mostly 0 and 10 values (0 meaning no missing data, 1 meaning 1 year of missing data, and so on).

So, I need to keep the countries whose rowSums value for this range is 0, 1, or 2. How would I integrate this condition into my code for the 'rowMeans' command so it only gives an output of countries missing 2 or less years (NA) of data in that time range?

Answer 1

With the rowSums create a logical index, use that to subset the data by specifying it as row index, get the rowMeans

i1 <- rowSums(is.na(countrydata[, 33:42])) <=2
Means <- rowMeans(countrydata[i1, 33:42], na.rm = TRUE)
country <- countrydata$country[i1]
data.frame(country, Means)

---

Or using tidyverse

library(dplyr)
countrydata %>%
    select(country, 33:42) %>%
    filter(rowSums(across(where(is.numeric), is.na)) <= 2) %>%
    transmute(country, Means = rowMeans(.[-1], na.rm = TRUE))