introducing n_Obs column to htest summary outputs

[IndrajeetPatil]

I am wondering if it will be a good idea to add a column to the summary outputs for htest objects containing sample size or number of observations information.

Although this might be straightforward for some tests, it's not so for others. Or at least I can't think of a straightforward way to extract this information in certain contexts.

# t-test ----------------------

mod1 <- t.test(mtcars$wt, mu = 3)
mod1$parameter[[1]] + 1
#> [1] 32

# but what about when var.equal = FALSE for two-sample test?
mod2 <- t.test(formula = wt ~ am, data = mtcars)
mod2$parameter[[1]] # sample size no longer parameter + 1
#> [1] 29.23352

# chi-squared test ------------

mod3 <- chisq.test(table(mtcars$am, mtcars$cyl))
#> Warning in chisq.test(table(mtcars$am, mtcars$cyl)): Chi-squared approximation
#> may be incorrect
sum(mod3$observed)
#> [1] 32

mod4 <- chisq.test(table(mtcars$cyl))
sum(mod4$observed)
#> [1] 32

# correlation ------------------

# no need, correlation package does this already
dplyr::select(mtcars, wt, mpg, drat) |>
  correlation::correlation() |>
  as.data.frame()
#>   Parameter1 Parameter2          r   CI     CI_low    CI_high         t
#> 1         wt        mpg -0.8676594 0.95 -0.9338264 -0.7440872 -9.559044
#> 2         wt       drat -0.7124406 0.95 -0.8499795 -0.4839784 -5.560807
#> 3        mpg       drat  0.6811719 0.95  0.4360484  0.8322010  5.096042
#>   df_error            p              Method n_Obs
#> 1       30 3.881876e-10 Pearson correlation    32
#> 2       30 9.568520e-06 Pearson correlation    32
#> 3       30 1.776240e-05 Pearson correlation    32

# etc.

^{Created on 2021-10-22 by the reprex package (v2.0.1)}

[vincentarelbundock]

FWIW, I don't think nobs is a parameter or something we estimate, so it should be out of scope for this package.

[bwiernik]

htest objects are such a mess, and there are too many cases where guesses would be wrong, that I think we don't do this

[vincentarelbundock]

Cool. Will close now to clean the repo up a bit, but @IndrajeetPatil should of course feel free to re-open if he has strong views or an implementation in mind.

[iago-pssjd]

I believe it would be a great idea, even not only that (/yet, better), but including the whole observed (and the expected?) table for a chi-squared test , like in summarytools::ctable (https://cran.r-project.org/web/packages/summarytools/vignettes/introduction.html#chi-square-%F0%9D%9B%982-odds-ratio-and-risk-ratio) or (with a more similar display) in compareGroups::createTable (https://cran.r-project.org/web/packages/compareGroups/vignettes/compareGroups_vignette.html#odds-ratios-and-hazard-ratios)

[strengejacke]

I think this is something we could realize in report, @rempsyc, what do you think?

[rempsyc]

Should be doable yeah