When using the option fill_gaps=True, the regions are all returned as "XX."

poverty <- povcalnet(year = 2015, fill_gaps = TRUE)
poverty <- povcalnet(year = 2015)

The povcalnet function uses readr::read_csv to format the data. By the default configuration for readr (version > 2.0), messages about the dataframe column types will pop up once the data is read. readr::read_csv offers an argument show_col_types to switch off the sometimes annoying messages. However, such an argument is not available when using povcalnet.

Currently, package internal variable are being defined at build time in add_internal_data.R
This is an issue for variables that are pulled from the API using povcalnet_info(). This approach breaks in the following cases:

• Information is updated on povcalnet PROD but the package is not being updated
• package is used on testing servers, but information is pulled from PROD

Hi,
The argument fill_gaps of povcalnet() is described as "logical: 'TRUE' will interpolate / extrapolate values when surveys are not available for a specific year." However, no interpolated result (poverty headcount/gini etc) is returned when this argument is TRUE. I requested data for all countries with fill_gaps==TRUE and got a returned dataframe with a variable isinterpolated. It turns out that isinterpolated is always 0 (indicates no interpolation), and the total amount of data is equal to that with fill_gaps==FALSE.

Is there a mistake in this function or something I have misunderstood?
Really need the interpolated data (or methodology). Thanks in advance.

Different coverage level mess up the aggregation behavior

For example this works:
povcalnet_cl(rep("CHN",76),seq(0.01,0.76,0.01),rep(2010,76))
while this:
povcalnet_cl(rep("CHN",77),seq(0.01,0.77,0.01),rep(2010,77))
produces an error:
"Error: assert_that: assertion must return a logical value"
[using a fully updated R version 3.6.1 (2019-07-05) on Ubuntu 18.04.3 LTS]

povcalnet(country = "XXX", year = 2002) throws an error. It should return a well formatted, empty data.frame

povcalnet_cl("ARG", 1.9, 1980) will not return any data because coverage is not specified, and the API only returns "national" coverage data by default.

TO DO:

• Make this more explicit in the function documentation
• Add helpful warnings when this happens

I am using povcalnetR to get the poverty headcounts for each country in a loop. However, the loop always gets stuck at some point such that I have to force shut down my R software. After checking the code, I found that certain countries caused the problem, e.g.:

library(povcalnetR)
povcalnet(
    country = "SUR", povline = 1.9, year = 2010:2021
  )

If country="all", the code works fine, and the retrieved data will not contain information for the problematic country "SUR". It seems that if a country is not in the database, the data retrieval for that individual country will raise a bug.

Other information:

• System: Windows 10
• R version: 4.1.1
• povcalnetR_0.1.1

try:
povcalnet("GNB",10,1991)
gives: -1 to headcount povertygap povertygapsq watts
it should give 1 to headcount at least

Hey @tonyfujs ,

I just checked and the only name of the function arguments that need to be changed is poverty_line= to pov_line=. In Stata the option name is povline()

Thanks.
Best,

Hi Tony,

By default, the Stata package loads coverages levels when option coverage is left empty povcalnet, country("all"). However, the R package filters everything that is not national by default. In this case, countries like Argentina, which does not have national coverage, are excluded from the output when povcalnet(country = "all") is executed.

I could update that if you want, but I thought it would be better if you do it.

Thanks.
Best,
Andres

The following lines of code will report a row with data for region SAS in 2015 but NAs for 2018.

library(povcalnetR)
regions <- povcalnet_wb()
regions <- regions[regions$regioncode == "SAS",]
regions[regions$year == 2015,]
regions[regions$year == 2018,]

The PovcalNet web interface, however, suppresses the results for both years as coverage in 2015 and 2018 in SAS is too low.

This inconsistency seems weird.

Ryan

In more detail the results are as follows (the rest of countries seem to return poverty rates for all survey years when poverty lines other than 1.9 $ are requested--not necessarily the same for each year):
[1] "PCN request for ARG returns an empty tibble (own PLs)."
[1] "PCN request for BOL returns a tibble with less years (own PLs). Missing: 1992"
[1] "PCN request for COL returns a tibble with less years (own PLs). Missing: 1980;1988;1989;1991"
[1] "PCN request for ECU returns a tibble with less years (own PLs). Missing: 1995"
[1] "PCN request for ETH returns a tibble with less years (own PLs). Missing: 1981"
[1] "PCN request for HND returns a tibble with less years (own PLs). Missing: 1986"
[1] "PCN request for URY returns a tibble with less years (own PLs). Missing: 1992;1995;1996;1997;1998;2000;2001;2002;2003;2004;2005"

It would appear that the povcalnet() function doesn't pull data for ARE. Even if I try using povcalnet(country = "ARE") it just returns an empty dataframe. Data exists on the povcal website so I'm not sure why the package ignores it.

Should be pulled automatically from pocalnet_info cache, not from a separate object -> Risk of being out of sync