CODEWITHSUNDEEP
rstatistics
SunnyShah
SunnyShah

Reputation: 30517

How prop.test(x,n) works internally?

> prop.test(6,10)

    1-sample proportions test with continuity correction

data:  6 out of 10, null probability 0.5
X-squared = 0.1, df = 1, p-value = 0.7518
alternative hypothesis: true p is not equal to 0.5
95 percent confidence interval:
    0.2736697 0.8630694
sample estimates:
    p 
0.6

Please help me with, How to calculate this manually?

Here is what I have tried,

> se =  sqrt(0.6*0.4*10/9)/sqrt(10)
> mean = 6/10
> z =  qnorm(0.975)
> from95 = mean - z*se
> to95 = mean + z*se
> from95
[1] 0.2799392
> to95
[1] 0.9200608

Upvotes: 0

Views: 809

Answers (1)

Rich Scriven
Rich Scriven

Reputation: 99361

All the information you need is in the source code of the function. You can separate it into parts with as.list(body(prop.test)). Or sending it to a file with sink so you can study it separately may help also.

For example, you can view the thirty-first element of the function body with

> as.list(body(prop.test))[[31]]
# if (alternative == "two.sided") PVAL <- pchisq(STATISTIC, PARAMETER, 
#     lower.tail = FALSE) else {
#     if (k == 1) 
#         z <- sign(ESTIMATE - p) * sqrt(STATISTIC)
#     else z <- sign(DELTA) * sqrt(STATISTIC)
#     PVAL <- pnorm(z, lower.tail = (alternative == "less"))
# }

Upvotes: 2

Related Questions