Statmanager-kr

<h2>Bug fix</h2>
<ul>
<li>I found and fixed some typos in the printed sentence when the language is set to &quot;eng&quot;.</li>
</ul>
<h2>Improvement</h2>
<ul>
<li>The result of the linear regression will now show the standardized regression coefficient beta as well as the unstandardized regression coefficient. This improvement is also available in Hierarchical Linear Regression and Multivariate Linear Regression.</li>
<li>The revised results table is shown below:</li>
</ul>

  | unstandadrized coefficient | standard error | standardized coefficient beta | p-value | 95% CI Low | 95% CI High
-- | -- | -- | -- | -- | -- | --

Improvement

- Now, the nominal variables provided as covariates in One-way ANCOVA and Repeated-Measures ANCOVA are automatically dummy-coded.
- Add docstrings for classes and methods.

---

개선

- ANCOVA 분석에서 투입된 공변량 중 명목변수들이 이제 자동으로 더미코딩됩니다.
- 각종 클래스와 메소드에 독스트링이 추가되었습니다.

Bug fix

I found a bug where if you run an analysis by applying a `selector` parameter in the `.progress()` method, and then immediately change the `selector` parameter again and run it once more, the filter is applied twice, rather than applying the new selector on the original data. This has now been fixed, and the `selector` works fine.

Improvement

In some analysis, I've changed the format of the result printed from `str` to `pd.DataFrame`, which should make it easier to see the results more clearly.

The following analyses are affected by this change:

- independent samples t-test
- dependent samples t-test
- Welch’s two sample t-test
- Yuen’s two sample t-test
- Mann-Whitney U test
- Brunner-Munzel Test
- Wilcoxon-Signed Rank Test
- Kruskal Wallis Test
- Friedman Test

---

버그 픽스

`.progress()` 메소드에서 `selector` 파라미터를 한 번 적용한 후, 다시 selector 파라미터를 변경하여 적용할 때 이중으로 필터가 걸리는 현상을 발견했습니다.

수정하였으며, 이제 정상 작동합니다.

개선

일부 분석에서 결과가 출력되는 방식을 `str`에서 `pd.DataFrame`으로 바꿨습니다.

이제 조금 더 보기 편할 것으로 보입니다.

해당 변경이 적용된 분석은 아래와 같습니다.

- independent samples t-test
- dependent samples t-test
- Welch’s two sample t-test
- Yuen’s two sample t-test
- Mann-Whitney U test
- Brunner-Munzel Test
- Wilcoxon-Signed Rank Test
- Kruskal Wallis Test
- Friedman Test

Improvement

Readability of tables printed as results in Linear or Logistic Regression were improved.

Columns were renamed, and DataFrame were restructured.

The mapping logic applied to the dependent variable in multinomial logistic regression is printed alongside.

See the examples as below:


  | Summary
-- | --
Model: | OLS
Dependent Variable: | postscore
Date: | 2024-01-10 15:08
No. Observations: | 30
Df Model: | 4
Df Residuals: | 25

Deprecating

`group_names` parameter

`group_names` parameter in `.progress()` was deprecated.

If you want to select `DataFrame`, use `selector` parameter please.

`FigureInStatmanager` Class

Now `FigureInStatmanager` class is deprecated.

All figures generated by running `.figure()` or .progress() is `matplotlib..axes.Axes` or `seaborn.FacetGrid`

This is probably more useful than writing complicated code to make it compatible.

Even for users who aren't familiar with the dependent libraries, a styled figure is returned by default.

Of course, if you're familiar with seaborn and matplotlib, you'll want to feel free to manipulate the properties of the returned object.

`group_names` 매개변수

.progress()`의` group_names` 파라미터는 더 이상 사용되지 않습니다.

특정 조건에 다라 DataFrame을 조정하려면 `selector` 파라미터를 대신 사용하시기 바랍니다.

`FigureInStatmanager` 클래스

이제 `FigureInStatmanager` 클래스는 더 이상 사용되지 않습니다.

.figure() 혹은 .progress9)에서 생성된 모든 그래프와 그림들은 이제 `matplotlib.Axes` 혹은 `seabron.FacetGrid` 객체로 반환됩니다.

모든 `matplotlib` 및 `seaborn` 메소드를 적용하여 그래프의 속성을 조절하길 바랍니다.

---

New analysis

**Hierarchical linear regression** is **available** now.

If you enter the `hier_linearr` in `method` parameter in `.progress()`, hierarchical linear regression will run.

The type of arguments entered in vars parameter in .progress() must be list. You can make a “step” distinct by providing a list as an element within a list. For example, if you want to make hierarchical linear regression model predicting ‘income’ by entering some variables, your code should look like this:

python
import pandas as pd
from statmanager import Stat_Manager

df = pd.read_csv(r"../..", index_col = 'id')
sm = Stat_Manager(df)

step_1 = ['age', 'sex', 'education'] ivs entered in step 1
step_2 = ['location', 'job', 'marriage'] ivs added in step 2 with step 1 variables

sm.progress(method = 'hier_linearr', vars = ['income', step_1, step_2])


Also, .figure() is avaiable. statmanager-kr will show residual plot of last regression model.

The result of hierarchical regression will show this :

  | added_vars | R-squared of Model | p-value of Model | R-squared increased | F | p-value of F
-- | -- | -- | -- | -- | -- | --
Step 1 | None | 0.209 | 0.192 | NaN | NaN | NaN
Step 2 | ….. | 0.222 | 0.270 | 0.013 | 0.401 | 0.533


  | Step 1 | Step 2
-- | -- | --
Model: | OLS | OLS
Dependent Variable: | income | income
Date: | 2024-01-05 11:53 | 2024-01-05 11:53
No. Observations: | 30 | 30
Df Model: | 4 | 5
Df Residuals: | 25 | 24

Improvements

For the purpose of applying analyses to reveal differences between groups, one common approach is to look at whether the normality assumptions are met in a cross-group dataset. Until now, it was not possible to apply the analysis for the purpose of testing the normality assumption implemented in statmanager-kr via the `group_vars` parameter.The `kstest`, `shapiro`, and `z_normal` methods now work perfectly when `group_vars` is provided. The results of the analysis to verify the normality assumption are now printed in the form of a `pandas.DataFrame`.

The `.figure()` function has been updated accordingly, and works fine in kstest. However, `shapiro` and `z_normal` still need more work. Please note that currently, `.figure()` does not work correctly **when a list with more than 3 elements is provided** in `group_vars`.

Other than that, I've fixed a bunch of small bugs that I found during my work. For example, I fixed an error that prevented the function from working if you provided a variable with missing values as the dependent variable in a regression analysis (`linearr` or `logisticr`).

Other than that, I've fixed a bunch of small bugs that I found during my work. For example, I fixed an error that prevented the function from working if you provided a variable with missing values as the dependent variable in a regression analysis (`linearr` or logisticr).

---

개선된 기능

집단 간 차이를 규명하기 위한 분석을 적용하기 위한 목적으로, 집단 간 데이터 세트에서 정규성 가정을 충족하는지 여부를 살펴보는 것은 일반적인 접근 방식 중 하나입니다. 지금까지는 정규성 가정의 충족 여부를 확인하기 위해 구현된 분석 방법들이 group_vars 파라미터가 제공될 경우 작동하지 않았습니다. 이제, 이러한 불편함이 개선되어 `kstest`, `shapiro` 및 `z_normal` 들이 `group_vars`를 제공했을 때에도 완벽하게 작동합니다. 이제 정규성 가정을 확인하기 위한 분석 결과들은 `pandas.DataFrame`의 형태로 제공됩니다.

이에 맞춰 .`figure()` 기능도 업데이트 되었고, `kstest`에서 정상적으로 작동합니다. 다만, `shapiro` 및 `z_normal`에서는 아직 더 수정이 필요합니다. 현재에는 `group_vars`에 **3개 이상의 요소를 가진 list가 제공될 경우** `.figure()`가 정상적으로 작동하지 않으니 유의하십시오.

그 외, 작업 중에 발견된 소소한 버그들이 수정되었습니다. 예를 들면, regression 분석에 종속 변수로 결측치가 포함된 변수를 제공하는 경우 함수가 작동하지 않던 오류를 고쳤습니다 (`linearr` or `logisticr`).