Fix paired t-test support for long-format data with grouped workflows

[Copilot]

nice_t_test() with paired = TRUE failed when used in dplyr::group_by() workflows with long-format data, throwing invalid type (list) for variable 'Pair(...)'. The function attempted to use R's formula method for paired tests, which is incompatible with R >= 4.4.0.

Changes

Core fix (R/nice_t_test.R)

• Extract numeric vectors directly for paired tests instead of creating formulas
• Pass vectors to t.test(x, y, paired = TRUE, ...) and effectsize::cohens_d(x, y, paired = TRUE, ...)
• Formula method now only used for independent samples tests

Tests (tests/testthat/test-nice_t_test.R)

• Added test coverage for paired t-tests in grouped long-format workflows

Documentation (man/nice_t_test.Rd)

• Clarified paired t-test usage and data format requirements
• Added paired t-test example

Version

• Bump 0.2.0 → 0.2.1 with NEWS.md entry

Example

Previously failing code now works:

df_long %>%
  pivot_longer(cols = c(quality, status), 
               names_to = "item", 
               values_to = "quality_status") %>%
  group_by(country, stimuli) %>%
  summarise(
    eff = nice_t_test(
      data = pick("quality_status", "item"), 
      response = "quality_status", 
      group = "item", 
      paired = TRUE
    )
  ) %>%
  unnest_wider(eff)

Results match the one-sample t-test workaround (computing differences manually) exactly.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Ask for help with nice_t_test() for paired samples t test</issue_title> <issue_description>Dear Dr. Thériault,

I hope this message finds you well.

Could you please help me with an issue I encountered when using nice_t_test() to run paired samples t tests for multiple groups? I’m currently using R 4.5.1 and rempsyc 0.2.0 on Windows OS.

I’m trying to perform paired t-tests for a set of stimuli, and I’ve attached my data for your reference.

df_long.xlsx

However, my code does not seem to work as intended. When I change it to a one-sample t-test, it runs successfully.

Paited t test did not work:

ttest <- df_long %>%
  pivot_longer(cols = c(quality, status), names_to = "item", values_to = "quality_status") %>% # prepare variable for paired samples t test
  dplyr::group_by(country, stimuli) %>%
  dplyr::summarise(eff = nice_t_test(data = pick("quality_status", "item"), response = "quality_status", group = "item", paired = TRUE)) %>%
  unnest_wider(eff)
pander(ttest)

One sample t test works:

ttest <- df_long %>%
  dplyr::mutate(quality_status = quality - status) %>% # prepare variable for one sample t test
  dplyr::group_by(country, stimuli) %>% 
  dplyr::summarise(eff = nice_t_test(data = pick("quality_status"), response = "quality_status", nu = 0)) %>%
  unnest_wider(eff)
pander(ttest)

Previously, I had a similar issue using t.test() for paired t-tests. While setting paired = TRUE worked in some cases, it sometimes produced incorrect results if the rows in my data were not properly ordered. I later realized that paired t-tests in R require a wide-format dataset. I’m wondering if this might also explain why nice_t_test() is not working with my long-format data.

Thank you very much for your time and help!

Best regards, Liman Wang </issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes rempsyc/rempsyc#81

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.