stepcount 3.7.1

What's Changed

• fix: use Int64Dtype, StringDtype to prevent int cast to float when NAs by @chanshing in #111

Full Changelog: v3.7.0...v3.7.1

stepcount 3.7.0

What's Changed

• feat: add --exclude-wear-below option to exlude days with insufficient wear time by @chanshing in #97
• feat: option --exclude-first-last to exclude first or last day, or both by @chanshing in #98
• ci: use artifacts to prevent http request rate limit exceeded by @chanshing in #99
• refactor: improve how ssl model is loaded by @chanshing in #100
• refactor(ssl): check model loaded first, then check len(X) by @chanshing in #101
• ci: use artifacts for model weights (the joblib files) for reuse by @chanshing in #103
• ci: remove install.yaml by @chanshing in #104
• feat(impute_missing): option to skip full missing days in imputation by @chanshing in #105
• tweaks to adjusted estimates by @chanshing in #106
• fix(skip_full_missing_days): fix bug all() -> any() by @chanshing in #107
• more flexible NA handling when adjusted estimation by @chanshing in #108
• change impute_days default median -> mean by @chanshing in #109

Full Changelog: v3.6.0...v3.7.0

stepcount 3.6.0

What's Changed

• Update README.md by @KPCOFGS in #92
• feat: field "Covers24hOK" to check 24h coverage by @chanshing in #95

New Contributors

• @KPCOFGS made their first contribution in #92

Full Changelog: v3.5.0...v3.6.0

stepcount 3.5.0

What's Changed

• chore: deprecate median ptiles at; shorten field names by @chanshing in #78
• fix: zero-out predicted steps < steptol by @chanshing in #79
• fix: impute walking separately by @chanshing in #80
• fix: use only walking windows when calculating max by @chanshing in #81
• Fix cadence's steptol: normalize to steps/min by @chanshing in #83
• Fix to use .agg(_aggfn) instead of .aggfn() to correctly handle NAs by @chanshing in #84
• Use coarser grid in imputation: 1-min -> 5-min grid by @chanshing in #85
• Fix cadence adjustment by @chanshing in #86
• Use .total_seconds() instead of .seconds; Fix .sum() of all-NaN to return NaN by @chanshing in #87
• chore(enmo): promptly downsample to reduce computation and memory by @chanshing in #89
• fix(enmo): use .agg(_aggfn) instead of .aggfn() to skipna properly by @chanshing in #90
• Concat ENMO and steps columns into one output dataframe/CSV by @chanshing in #91

Full Changelog: v3.4.0...v3.5.0

stepcount 3.4.0

What's Changed

• feat: 95th percentile of cadence by @chanshing in #69
• fix(predict): handle zero-length X inputs by @chanshing in #70
• fix NoneType error by @chanshing in #71
• chore: silence warning 'Mean of empty slice' by @chanshing in #72
• feat: steps 5th & 95th ptiles at; refactor field names by @chanshing in #74
• fix: use only walking windows when calculating 95th cadence by @chanshing in #75
• feat: output minute-level data by @chanshing in #76
• feat: output enmo metrics by @chanshing in #77

Full Changelog: v3.3.1...v3.4.0

stepcount 3.3.1

What's Changed

• fix: peak cadence to nan when not enough data by @chanshing in #68

Full Changelog: v3.3.0...v3.3.1

stepcount 3.3.0

What's Changed

• feat: support specifying custom headers for CSV inputs by @chanshing in #66

Full Changelog: v3.2.5...v3.3.0

stepcount 3.2.5

What's Changed

• refactor: use mean for infer_freq but remove outliers by @chanshing in #64
• fix NaT freq issue when recording is extremely small by @chanshing in #65

Full Changelog: v3.2.4...v3.2.5

stepcount 3.2.4

What's Changed

• fix(hmm_utils): update module name MultinomialHMM -> CategoricalHMM after hmmlearn 0.3.0 by @chanshing in #63

Full Changelog: v3.2.3...v3.2.4

stepcount 3.2.3

What's Changed

• fix(read): handle filenames containing dots, confusing extension by @chanshing in #60

Full Changelog: v3.2.2...v3.2.3