Testing - tern fix for analysis function refactor bugs

[edelarua]

See insightsengineering/tern#1406

[github-actions]

Unit Tests Summary

  1 files  111 suites   3m 28s ⏱️
245 tests 245 ✅ 0 💤 0 ❌
522 runs  522 ✅ 0 💤 0 ❌

Results for commit 9ba1743.

♻️ This comment has been updated with latest results.

[github-actions]

Unit Test Performance Difference

Test Suite	$Status$	Time on main	$±Time$	$±Tests$	$±Skipped$	$±Failures$	$±Errors$
pagination_table	💔	$12.21$	$+1.99$	$+9$	$-6$	$0$	$0$
table_aet03	💔	$1.58$	$+3.33$	$+4$	$-1$	$0$	$0$
table_aet04	💔	$6.41$	$+5.83$	$+14$	$-8$	$0$	$0$
table_ent	💚	$13.98$	$-1.07$	$0$	$-4$	$0$	$0$
table_pkct01	💔	$1.55$	$+1.76$	$+6$	$-2$	$0$	$0$

Additional test case details

Test Suite	$Status$	Time on main	$±Time$	Test Case
pagination_table	💔	$1.55$	$+2.11$	Pagination_works_for_page_types
table_aet03	💔	$1.58$	$+3.33$	AET03_variant_1_is_produced_correctly
table_pkct01	💔	$1.20$	$+1.55$	Specific_PKCT01_features_are_present

Results for commit ddac8b7

♻️ This comment has been updated with latest results.

[edelarua]

The snapshot update is not related to tern - it comes from the data.

I'm not sure when the data was changed, but the change to the LBT03 snapshot reflects the expected result as all CHG values are NA when AVISIT == "Baseline", so this change should be merged in.

pharmaverseadam::adlb |> 
  dplyr::filter(AVISIT == "Baseline") |> 
  dplyr::pull(CHG) |> 
  unique()
#> [1] NA

^{Created on 2025-03-06 with reprex v2.1.1}

[shajoezhu]

The snapshot update is not related to tern - it comes from the data.

I'm not sure when the data was changed, but the change to the LBT03 snapshot reflects the expected result as all CHG values are NA when AVISIT == "Baseline", so this change should be merged in.

pharmaverseadam::adlb |> 
  dplyr::filter(AVISIT == "Baseline") |> 
  dplyr::pull(CHG) |> 
  unique()
#> [1] NA

Created on 2025-03-06 with reprex v2.1.1

this is really odd. lets follow up, @edelarua , I was wondeing when updating the snapshot, are you updating it locally on your machine?

[edelarua]

The snapshot update is not related to tern - it comes from the data.
I'm not sure when the data was changed, but the change to the LBT03 snapshot reflects the expected result as all CHG values are NA when AVISIT == "Baseline", so this change should be merged in.

pharmaverseadam::adlb |> 
  dplyr::filter(AVISIT == "Baseline") |> 
  dplyr::pull(CHG) |> 
  unique()
#> [1] NA

Created on 2025-03-06 with reprex v2.1.1

this is really odd. lets follow up, @edelarua , I was wondeing when updating the snapshot, are you updating it locally on your machine?

Yes, it happens locally and in the integration tests. The updated results make sense since the change from baseline values at the baseline visit should be 0/NA, not what we had previously.

[shajoezhu]

The snapshot update is not related to tern - it comes from the data.
I'm not sure when the data was changed, but the change to the LBT03 snapshot reflects the expected result as all CHG values are NA when AVISIT == "Baseline", so this change should be merged in.

pharmaverseadam::adlb |> 
  dplyr::filter(AVISIT == "Baseline") |> 
  dplyr::pull(CHG) |> 
  unique()
#> [1] NA

Created on 2025-03-06 with reprex v2.1.1

this is really odd. lets follow up, @edelarua , I was wondeing when updating the snapshot, are you updating it locally on your machine?

Yes, it happens locally and in the integration tests. The updated results make sense since the change from baseline values at the baseline visit should be 0/NA, not what we had previously.

I would like to check the computation in tern again, perhaps we caught a bug that was not found before. The data change happened for some time already, but we have been refactoring the computation.

[edelarua]

Hi @shajoezhu & @Melkiades,

I have pinpointed where exactly the change is coming from: https://github.com/insightsengineering/tern/pull/1401/files#diff-19af54bdeb1d75fc74da8834bca1f4d2c90d1d33c3650aa93bab0fae8505a7ceR201

The inclNA argument of summarize_change() was previously set (incorrectly) to na_rm and I changed it to !na_rm since including NAs is the same as not removing NAs. This was originally added in this PR - I remember catching this when reviewing the analyze_vars() function as well.

This is why the snapshot changed here and then back to the original result in this PR.

I think in light of this we can merge insightsengineering/tern#1406 and then this change in.

[shajoezhu]

Thank you so much @edelarua ! this is good!

[shajoezhu]

lgtm! thanks @edelarua