Additional Options

Formatting umncvmstats Results

All output from statistical tests in this package are stored as a special object (an atest) that knows how to print itself in a nice way.

It does this by converting it into a gt object from the gt package.

You can use any gt function by first doing this conversion yourself using as_gt. If you use any gt functions you’ll need to load the gt library before umncvmstats.

The following code simply converts to a gt object, though has no change from the default output because I haven’t added any other functions yet.

a <- pairwise_t_inference(mpg ~ cyl, data=mtcars2)
a |> as_gt()

response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value	p.adjust
mpg	cyl: 4 - 6	6.9	1.5	13.0	2.9	10.9	0.000	4.72	0.0004	0.0012
mpg	cyl: 4 - 8	11.6	1.5	15.0	7.5	15.7	0.000	7.60	< 0.0001	< 0.0001
mpg	cyl: 6 - 8	4.64	0.88	18.5	2.33	6.95	0.000	5.29	< 0.0001	0.0001
Welch Two Sample t-test (two.sided), with 95% confidence intervals, adjusted for 3 comparisons using the Bonferroni method.
p-values adjusted for 3 multiple comparisons using the Bonferroni method.

Compacting Results

The tab_compact function makes the table of results smaller; by default it uses font.size = 13 and padding = 1.

a |> as_gt() |> tab_compact()

response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value	p.adjust
mpg	cyl: 4 - 6	6.9	1.5	13.0	2.9	10.9	0.000	4.72	0.0004	0.0012
mpg	cyl: 4 - 8	11.6	1.5	15.0	7.5	15.7	0.000	7.60	< 0.0001	< 0.0001
mpg	cyl: 6 - 8	4.64	0.88	18.5	2.33	6.95	0.000	5.29	< 0.0001	0.0001
Welch Two Sample t-test (two.sided), with 95% confidence intervals, adjusted for 3 comparisons using the Bonferroni method.
p-values adjusted for 3 multiple comparisons using the Bonferroni method.

For comparison, here is font.size = 11 and padding = 3.

a |> as_gt() |> tab_compact(font.size = 11, padding = 3)

response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value	p.adjust
mpg	cyl: 4 - 6	6.9	1.5	13.0	2.9	10.9	0.000	4.72	0.0004	0.0012
mpg	cyl: 4 - 8	11.6	1.5	15.0	7.5	15.7	0.000	7.60	< 0.0001	< 0.0001
mpg	cyl: 6 - 8	4.64	0.88	18.5	2.33	6.95	0.000	5.29	< 0.0001	0.0001
Welch Two Sample t-test (two.sided), with 95% confidence intervals, adjusted for 3 comparisons using the Bonferroni method.
p-values adjusted for 3 multiple comparisons using the Bonferroni method.

Rounding umncvmstats Results

By significant digits or decimals

It’s trickier than I think it should be to get R to output a reasonable number of digits; this was one reason I wrote this package.

The formatted output from statistical inference functions in this package makes a guess that should be reasonable in most, but not all, cases.

To modify the number of digits that are output, use the set_digits function. You can either specify the number of significant digits or the number of decimal places. By default, digits for the estimate, the SE, and the confidence bounds are set for all rows.

The set_digits function should be used directly on the results object, before any gt formatting.

The code sets the number of significant digits to 2 (for the estimate, the SE, and the confidence bounds),

a |> set_digits(digits=2)

response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value	p.adjust
mpg	cyl: 4 - 6	6.9	1.5	13.0	2.9	11	0.000	4.72	0.0004	0.0012
mpg	cyl: 4 - 8	12	1.5	15.0	7.5	16	0.000	7.60	< 0.0001	< 0.0001
mpg	cyl: 6 - 8	4.6	0.88	18.5	2.3	7.0	0.000	5.29	< 0.0001	0.0001
Welch Two Sample t-test (two.sided), with 95% confidence intervals, adjusted for 3 comparisons using the Bonferroni method.
p-values adjusted for 3 multiple comparisons using the Bonferroni method.

This code sets the number of decimal places to 2.

a |> set_digits(decimals=2)

response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value	p.adjust
mpg	cyl: 4 - 6	6.92	1.47	13.0	2.89	10.95	0.000	4.72	0.0004	0.0012
mpg	cyl: 4 - 8	11.56	1.52	15.0	7.46	15.67	0.000	7.60	< 0.0001	< 0.0001
mpg	cyl: 6 - 8	4.64	0.88	18.5	2.33	6.95	0.000	5.29	< 0.0001	0.0001
Welch Two Sample t-test (two.sided), with 95% confidence intervals, adjusted for 3 comparisons using the Bonferroni method.
p-values adjusted for 3 multiple comparisons using the Bonferroni method.

Specific values can be specified using the column names and row numbers.

a |> set_digits(decimals=5, columns="difference", rows=1:2)

response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value	p.adjust
mpg	cyl: 4 - 6	6.92078	1.5	13.0	2.9	10.9	0.000	4.72	0.0004	0.0012
mpg	cyl: 4 - 8	11.56364	1.5	15.0	7.5	15.7	0.000	7.60	< 0.0001	< 0.0001
mpg	cyl: 6 - 8	4.64	0.88	18.5	2.33	6.95	0.000	5.29	< 0.0001	0.0001
Welch Two Sample t-test (two.sided), with 95% confidence intervals, adjusted for 3 comparisons using the Bonferroni method.
p-values adjusted for 3 multiple comparisons using the Bonferroni method.

You can also set the number of decimals based on significant digits for a given column or columns; for example, when created, the decimals are set based on the number of significant digits needed for the SE and ME columns, where ME (the margin of error) is calculated from conf.high and conf.low.

This determines how many decimals are needed to display the SE and ME with 2 significant digits, and uses that for the estimate, the SE, and the confidence bounds. The following is the current default setting.

a |> set_digits(digits=2, by=c("SE", "ME"))

response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value	p.adjust
mpg	cyl: 4 - 6	6.9	1.5	13.0	2.9	10.9	0.000	4.72	0.0004	0.0012
mpg	cyl: 4 - 8	11.6	1.5	15.0	7.5	15.7	0.000	7.60	< 0.0001	< 0.0001
mpg	cyl: 6 - 8	4.64	0.88	18.5	2.33	6.95	0.000	5.29	< 0.0001	0.0001
Welch Two Sample t-test (two.sided), with 95% confidence intervals, adjusted for 3 comparisons using the Bonferroni method.
p-values adjusted for 3 multiple comparisons using the Bonferroni method.

Rounding p-values

Rounding p-values are currently handled within the gt mechanism, using the fmt_pvalue function, which formats p-values (anything named p.value or p.adjust), with digits significant digits, up to max.digits decimals. The default is 2 significant digits, and 4 maximum digits.

Here’s what the output would look like with 3 significant digits and 5 maximum digits.

a |> as_gt() |> fmt_pvalue(digits=3, max.digits=5)

response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value	p.adjust
mpg	cyl: 4 - 6	6.9	1.5	13.0	2.9	10.9	0.000	4.72	0.00040	0.00121
mpg	cyl: 4 - 8	11.6	1.5	15.0	7.5	15.7	0.000	7.60	< 0.00001	< 0.00001
mpg	cyl: 6 - 8	4.64	0.88	18.5	2.33	6.95	0.000	5.29	0.00005	0.00014
Welch Two Sample t-test (two.sided), with 95% confidence intervals, adjusted for 3 comparisons using the Bonferroni method.
p-values adjusted for 3 multiple comparisons using the Bonferroni method.

Output as data set

To get the underlying output as a data set, use as_tibble. You’ll need to have loaded the tibble library, which loads as part of the tidyverse. There are several options for how to handle the footnotes.

This is useful when further processing or graphics based on output is desired.

You’ll notice that the column names are different from the formatted output; the formatted output tries to simplify the table to make it clearer to read, while this instead returns the underlying pieces for you to handle as needed in your specific case.

a |> as_tibble()
#> # A tibble: 3 × 13
#>   .y    .x    .x_contrast difference    SE    df conf.low conf.high  null
#>   <chr> <chr> <chr>            <dbl> <dbl> <dbl>    <dbl>     <dbl> <dbl>
#> 1 mpg   cyl   4 - 6             6.92 1.47   13.0     2.89     10.9      0
#> 2 mpg   cyl   4 - 8            11.6  1.52   15.0     7.46     15.7      0
#> 3 mpg   cyl   6 - 8             4.64 0.877  18.5     2.33      6.95     0
#> # ℹ 4 more variables: t.value <dbl>, p.value <dbl>, p.adjust <dbl>, about <chr>

Multiple tests

In the one and two-sample tests (that is, when you’re not fitting a model), you can use + to specify that the test should be done on several variables.

one_t_inference(wt + mpg ~ am + vs, data = mtcars2)

response	variable	n	mean	SE	df	conf.low	conf.high
wt	am = automatic	19	3.77	0.18	18	3.39	4.14
wt	am = manual	13	2.41	0.17	12	2.04	2.78
wt	vs = V-shaped	18	3.69	0.21	17	3.24	4.14
wt	vs = straight	14	2.61	0.19	13	2.20	3.02
mpg	am = automatic	19	17.15	0.88	18	15.30	19.00
mpg	am = manual	13	24.4	1.7	12	20.7	28.1
mpg	vs = V-shaped	18	16.62	0.91	17	14.70	18.54
mpg	vs = straight	14	24.6	1.4	13	21.5	27.7
One Sample t-test (two.sided), with 95% confidence intervals.

two_t_inference(wt + mpg ~ am + vs, data = mtcars2)

response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value
wt	am: automatic - manual	1.36	0.25	29.2	0.85	1.86	0.000	5.49	< 0.0001
wt	vs: V-shaped - straight	1.08	0.29	30.0	0.49	1.66	0.000	3.76	0.0007
mpg	am: automatic - manual	−7.2	1.9	18.3	−11.3	−3.2	0.000	-3.77	0.0014
mpg	vs: V-shaped - straight	−7.9	1.7	22.7	−11.5	−4.4	0.000	-4.67	0.0001
Welch Two Sample t-test (two.sided), with 95% confidence intervals.

You can also use | to split the data set into groups before running the tests.

one_t_inference(wt + mpg ~ am | cyl, data = mtcars2)

	response	variable	n	mean	SE	df	conf.low	conf.high
cyl = 4	wt	am = automatic	3	2.94	0.24	2	1.92	3.95
	wt	am = manual	8	2.04	0.14	7	1.70	2.38
	mpg	am = automatic	3	22.90	0.84	2	19.29	26.51
	mpg	am = manual	8	28.1	1.6	7	24.3	31.8
cyl = 6	wt	am = automatic	4	3.389	0.058	3	3.204	3.574
	wt	am = manual	3	2.755	0.074	2	2.437	3.073
	mpg	am = automatic	4	19.12	0.82	3	16.53	21.72
	mpg	am = manual	3	20.57	0.43	2	18.70	22.43
cyl = 8	wt	am = automatic	12	4.10	0.22	11	3.62	4.59
	wt	am = manual	2	3.37	0.20	1	0.83	5.91
	mpg	am = automatic	12	15.05	0.80	11	13.29	16.81
	mpg	am = manual	2	15.40	0.40	1	10.32	20.48
One Sample t-test (two.sided), with 95% confidence intervals.

two_t_inference(wt + mpg ~ am | cyl, data = mtcars2)

	response	variable	difference	SE	df	conf.low	conf.high	null	t.value	p.value
cyl = 4	wt	am: automatic - manual	0.89	0.28	3.65	0.10	1.69	0.000	3.23	0.036
	mpg	am: automatic - manual	−5.2	1.8	9.00	−9.2	−1.1	0.000	-2.89	0.018
cyl = 6	wt	am: automatic - manual	0.634	0.094	4.17	0.377	0.891	0.000	6.74	0.0022
	mpg	am: automatic - manual	−1.44	0.92	4.41	−3.92	1.03	0.000	-1.56	0.19
cyl = 8	wt	am: automatic - manual	0.73	0.30	4.37	−0.07	1.54	0.000	2.46	0.065
	mpg	am: automatic - manual	−0.35	0.90	10.2	−2.34	1.64	0.000	-0.391	0.70
Welch Two Sample t-test (two.sided), with 95% confidence intervals.

Combining output

You can use combine_tests to combine output from several tests together. The package will do its best to output this meaningfully, but it’s up to you to combine meaningful things together.

combine_tests(
  one_t_inference(wt ~ am|vs, data = mtcars2),
  two_t_inference(wt ~ am, data = mtcars2),
  correlation_inference(mpg ~ wt + hp, data=mtcars2),
  one_proportion_inference(vs ~ am|cyl, data=mtcars2)) |>
  as_gt() |> tab_compact()

	response	variable	n	x	mean	difference	correlation	proportion	SE	df	conf.low	conf.high	null	t.value	p.value	footnote
vs = V-shaped	wt	am = automatic	12		4.10				0.22	11.0	3.62	4.59				1
	wt	am = manual	6		2.86				0.20	5.00	2.35	3.37				1
vs = straight	wt	am = automatic	7		3.19				0.13	6.00	2.87	3.52				1
	wt	am = manual	7		2.03				0.17	6.00	1.62	2.44				1
NA	wt	am: automatic - manual				1.36			0.25	29.2	0.85	1.86	0.000	5.49	< 0.0001	2
	mpg	wt					−0.868				−0.934	−0.744			< 0.0001	3
	mpg	hp					−0.78				−0.89	−0.59			< 0.0001	3
cyl = 4	vs = straight	am = automatic	3	3				1.00			0.29	1.00				4,5
	vs = straight	am = manual	8	7				0.88	0.12		0.47	1.00				4,5
cyl = 6	vs = straight	am = automatic	4	4				1.00			0.40	1.00				4,5
	vs = straight	am = manual	3	0				0.00			0.00	0.71				4,5
cyl = 8	vs = straight	am = automatic	12	0				0.00			0.00	0.26				4,6
	vs = straight	am = manual	2	0				0.00			0.00	0.84				4,5
1 One Sample t-test (two.sided), with 95% confidence intervals.
2 Welch Two Sample t-test (two.sided), with 95% confidence intervals.
3 Pearson's product-moment correlation (two.sided), with 95% confidence intervals
4 Exact binomial test (two.sided), with 95% confidence intervals.
5 Method chosen because n < 10.
6 Method chosen because observed proportion < 0.10.