File talk:Religious affiliation in New Zealand 1991-2006.svg

Source code and data

R code (offered under the GPL v2 or later license, as well as the usual CC-BY-SA and GFDL licenses):

library(ggplot2)
rel2 <- read.csv("NZ religious affiliation 1991-2006.csv")
svg("Religious affiliation in New Zealand 1991-2006.svg", width=5.6, height=3.5)
p <- ggplot(rel2, aes(x=Year, y=Percentage, group=Religion))
p + geom_line(aes(colour = Religion)) + geom_point(aes(colour = Religion, shape = Religion)) + 
  scale_x_continuous(lim = c(1991, 2006), breaks=c(1991, 1996, 2001, 2006))
dev.off()

Data, stored in "NZ religious affiliation 1991-2006.csv":

"Religion","Percentage","Year"
"Christian",70.22794,1991
"No Religion",20.2187572,1991
"Object to Answering",7.5907304,1991
"Other",1.96257,1991
"Christian",64.2951,1996
"No Religion",25.4680196,1996
"Object to Answering",7.5350938,1996
"Other",2.70178,1996
"Christian",60.6,2001
"No Religion",29.636936,2001
"Object to Answering",6.897005,2001
"Other",2.86606,2001
"Christian",55.6,2006
"No Religion",34.6480923,2006
"Object to Answering",6.4805703,2006
"Other",3.27134,2006

Source: Statistics New Zealand. For links see image page.

-- Avenue (talk) 12:32, 13 January 2010 (UTC)

Others calculation

Good work on the graph. Sorry but I think the Others subtraction method has a fault, please consider for example:

• 60% say Christian
• 35% say No religion
• 5% say Refuse
• 4% say Other

Then 100% - (60 + 35 + 5) = 0% for others. The method would be ok prior to 2001 when only one response was taken. For Others 2001, 2006, I think the numbers ought to be 4.225 and 5.447, adding Buddhist + Hindu + Islam, etc

The answered more than one doesn't matter here - 98% of Kiwis speak English, 4.2% Māori. Benchill (talk) 23:28, 13 January 2010 (UTC)

Thanks for the compliment, and your thoughts. I agree the current numbers are flawed, because people who say they are Christian as well as an "Other" religion are only counted under "Christian". However I think it's not quite as simple as you imply, because your "Other" numbers are also flawed. They double-count people who say they affiliate with two different "Other" religions. So my numbers will underestimate the true "Other" figures, and yours will overestimate. Without some information about the overlaps, we can't know which approach is more accurate. But I prefer my numbers, for two reasons:

1. the figures from your approach depend on the details of which religions within the "Other" group are reported by Stats NZ, and
2. I suspect the overlaps within the "Other" religions are greater than the overlaps between "Other" and "Christian".

Another approach would be to try to come up with a better estimate, based on some assumptions about the overlaps, but that could be seen as original research. Of course, the best approach would be to get the true "Other" figures from Stats NZ. -- Avenue (talk) 02:43, 14 January 2010 (UTC)

More about your example: I think there shouldn't be any overlap between "Object to Answer" and the various religions, nor between "No religion" and actual religions. So 35%+5%=40% of people would not have given any religion, and 60% of people were Christian, so all people with religions were Christian. Thus all the people with "Other" religions were also Christian, i.e. there is complete overlap. This is the worst possible case for my approach. While it is theoretically possible, I don't think it's very realistic. -- Avenue (talk) 02:51, 14 January 2010 (UTC)

You're right, the person for example ticking two boxes within others shouldn't be counted twice. Benchill (talk) 07:44, 14 January 2010 (UTC)