[R] Setting up hypothesis tests with the infer library?

[Rui Barradas]

Às 19:19 de 29/03/2025, Ebert,Timothy Aaron escreveu:
> How about calculating a 95% confidence interval about the estimated proportion in favor. The PooledInfRate package will do this for you. If confidence intervals overlap then there is no significant difference.
> 
> -----Original Message-----
> From: R-help <r-help-bounces using r-project.org> On Behalf Of Kevin Zembower via R-help
> Sent: Saturday, March 29, 2025 12:10 PM
> To: R-help email list <r-help using r-project.org>
> Subject: [R] Setting up hypothesis tests with the infer library?
> 
> [External Email]
> 
> Hello, all,
> 
> We're now starting to cover hypothesis tests in my Stats 101 course. As usual in courses using the Lock5 textbook, 3rd ed., the homework answers are calculated using their StatKey application. In addition (and for no extra credit), I'm trying to solve the problems using R. In the case of hypothesis test, in addition to manually setting up randomized null hypothesis distributions and graphing them, I'm using the infer library. I've been really impressed with this library and enjoy solving this type of problem with it.
> 
> One of the first steps in solving a hypothesis test with infer is to set up the initial sampling dataset. Often, in Lock5 problems, this is a dataset that can be downloaded with library(Lock5Data). However, other problems are worded like this:
> 
> ===========================
> In 1980 and again in 2010, a Gallup poll asked a random sample of 1000 US citizens "Are you in favor of the death penalty for a person convicted of murder?" In 1980, the proportion saying yes was 0.66. In 2010, it was 0.64. Does this data provide evidence that the proportion of US citizens favoring the death penalty was higher in 1980 than it was in 2010? Use p1 for the proportion in 1980 and p2 for the proportion in 2010.
> ============================
> 
> I've been setting up problems like this with code similar to:
> ===========================
> df <- data.frame(
>      survey = c(rep("1980", 1000), rep("2010", 1000)),
>      DP = c(rep("Y", 0.66*1000), rep("N", 1000 - (0.66*1000)),
>             rep("Y", 0.64*1000), rep("N", 1000 - (0.64*1000))))
> 
> (d_hat <- df %>%
>       specify(response = DP, explanatory = survey, success = "Y") %>%
>       calculate(stat = "diff in props", order = c("1980", "2010"))) ============================
> 
> My question is, is this the way I should be setting up datasets for problems of this type? Is there a more efficient way, that doesn't require the construction of the whole sample dataset?
> 
> It seems like I should be able to do something like this:
> =================
> (df <- data.frame(group1count = 660, #Or, group1prop = 0.66
>                   group1samplesize = 1000,
>                   group2count = 640, #Or, group2prop = 0.64
>                   group2samplesize = 1000)) =================
> 
> Am I overlooking a way to set up these sample dataframes for infer?
> 
> Thanks for your advice and guidance.
> 
> -Kevin
> 
> 
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide https://www.r-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide https://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
Hello,

Package PooledInfRate seems promising.


library(PooledInfRate)
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

year <- c(1980, 2010)
p <- c(0.66, 0.64)
n <- c(1000, 1000)
df1 <- data.frame(year, p, n)

df1 %>%
   mutate(yes = p * n, no = n - yes) %>%
   select(-p, -n) %>%
   tidyr::pivot_longer(-year, names_to = "answer", values_to = "counts") %>%
   mutate(answer = as.integer(answer == "yes")) %>%
   pooledBin(answer ~ counts | year, data = .)
#>   year           P        Lower       Upper
#> 1 1980 0.001113165 6.831431e-05 0.009052399
#> 2 2010 0.001102918 6.796757e-05 0.008755895


The CI's intersect, so support for the death penalty hasn't changed from 
1980 to 2010.

Hope this helps,

Rui Barradas




-- 
Este e-mail foi analisado pelo software antivírus AVG para verificar a presença de vírus.
www.avg.com