# Resampling Techniques Library for Statistical measure
Library for resampling methods over Chi-Abs, correlation, power analysis, linear regression (beta value), relative risk -- LS 40 @UCLA Life science course
RESAMP Library Documentation
- Author: Vishanth Hari Raj
- GitHub Repository: https://github.com/vishanth10/resamp
`resamp.py` Version 1.6.8 Documentation
Overview
`resamp.py` is a comprehensive Python library designed for statistical analysis and resampling techniques. This document serves as a guide for utilizing the library's functions, including statistical tests, data analysis, and visualization tools.
Installation
To install `resamp.py`, run the following command in your terminal:
```bash
pip install statistics_library
```
Functions:
Median Absolute Deviation (MAD)
`median_absolute_deviation(data)`
Calculates the Median Absolute Deviation of a dataset.
- **Parameters**:
- `data` (_array-like or str_): A 1D array or list containing the dataset, or a filename pointing to a CSV or Excel file containing a single column of numbers.
- **Returns**:
- _float_: The Median Absolute Deviation of the dataset.
- **Usage Example**:
```python
mad = median_absolute_deviation([1, 2, 3, 4, 5])
print(mad)
```
**Chi-Squared Test**
`calculate_chi_squared(observed, expected)`
Calculates the chi-squared statistic given observed and expected frequencies.
- **Parameters**:
- `observed` (_DataFrame_): Observed frequencies.
- `expected` (_DataFrame_): Expected frequencies.
- **Returns**:
- _float_: The chi-squared statistic.
Bootstrap Chi Absolute
`bootstrap_chi_abs(observed_data, num_simulations=10000, with_replacement=True)`
Generates a bootstrap distribution of the chi absolute statistic for an n*n contingency table.
- **Parameters**:
- `observed_data` (_np.array or pd.DataFrame_): Contingency table with observed frequencies.
- `num_simulations` (_int_): Number of bootstrap samples to generate.
- `with_replacement` (_bool_): Indicates whether sampling should be with replacement.
- **Returns**:
- _np.array_: Bootstrap distribution of chi absolute values.
**Relative Risk Analysis**
`calculate_relative_risk_two_treatments(observed_data, event_row_index, treatment1_index, treatment2_index)`
Calculates the relative risk of an event between two treatments.
- **Parameters**:
- `observed_data` (_np.array_): The observed data as a 2D array.
- `event_row_index` (_int_): Row index for the event.
- `treatment1_index` (_int_): Column index for the first treatment.
- `treatment2_index` (_int_): Column index for the second treatment.
- **Returns**:
- _float_: The relative risk.
-
**Correlation Analysis**
`permute_correlation(x, y, num_simulations=10000)`
Generates simulated correlation coefficients by permuting one variable.
- **Parameters**:
- `x` (_np.array_): Values of variable 1.
- `y` (_np.array_): Values of variable 2.
- `num_simulations` (_int_): Number of permutations.
- **Returns**:
- _np.array_: Simulated correlation coefficients.
**Linear Regression Analysis**
`bootstrap_confidence_interval(x, y, n_bootstrap=1000, confidence_level=99, return_type='both')`
Calculates bootstrap confidence intervals for the slope and intercept of a linear regression model.
- **Parameters**:
- `x`, `y` (_np.array_): Predictor and response variables.
- `n_bootstrap` (_int_): Number of bootstrap samples.
- `confidence_level` (_float_): Confidence level for the interval.
- `return_type` (_str_): Determines if the output is for 'slope', 'intercept', or 'both'.
- **Returns**:
- _dict_: Confidence intervals for the slope and/or intercept.
**Power Analysis**
`power_analysis(...)`
Performs a power analysis to determine the required sample size for achieving a specified power level.
- **Parameters**:
- `obs_diff`: the observed difference in medians or means between the two groups, depending on 'measure'.
- `group1`: data for group 1.
- `group2`: data for group 2.
- `num_simulations`: number of simulations to perform.
- `alpha`: significance level.
- `power_threshold`: desired power level to achieve.
- `factor_limit`: the maximum factor by which to increase the sample size.
- `measure`: 'median' or 'mean', the statistical measure to use for comparison.
- `verbose`: if True, print intermediate results.
- **Returns**:
- `required_sample_sizes`: a tuple of the required sample sizes for group 1 and group 2 to achieve the desired power.
- `achieved_power`: the power that was achieved with the returned sample sizes.
### Usage Example for `power_analysis`
```python
group1 = np.random.normal(5, 2, 100)
group2 = np.random.normal(5.5, 2, 100)
sample_sizes, achieved_power = power_analysis(obs_diff=0.5, group1=group1, group2=group2, alpha=0.05, power_threshold=0.8)
print(f'Required Sample Sizes: {sample_sizes}, Achieved Power: {achieved_power}')
```
**#Function Usage**
```python
def mean_absolute_deviation(data):
# Your implementation here
```
Calculates the mean absolute deviation of the provided dataset. Supports direct list input or file paths.
```python
def calculate_chi_squared(observed, expected):
# Your implementation here
```
Calculates the chi-squared statistic based on observed and expected frequencies.
```python
def calculate_expected(observed):
# Your implementation here
```
Calculates expected frequencies for a chi-squared test from observed data.
```python
def bootstrap_chi_abs(observed_data, num_simulations=10000, with_replacement=True):
# Your implementation here
```
Generates bootstrap samples for chi absolute statistics and calculates the chi absolute statistic for each sample.
```python
def calculate_relative_risk_two_treatments(observed_data, event_row_index, treatment1_index, treatment2_index):
# Your implementation here
```
Calculates the relative risk between two treatments.
```python
def permute_correlation(x, y, num_simulations=10000):
# Your implementation here
```
Performs permutation tests to calculate the null distribution of the correlation coefficient.
```python
def bootstrap_confidence_interval(x, y, n_bootstrap=1000, confidence_level=99, return_type='both'):
# Your implementation here
```
Calculates bootstrap confidence intervals for linear regression parameters.
```python
def plot_bootstrap_lines(x, y, n_bootstrap=1000, original_slope=2, original_intercept=0):
# Your implementation here
```
Plots regression lines for bootstrap samples and compares them with the original regression line.
```python
def power_analysis(...):
# Your implementation here
```
Performs a power analysis to determine the required sample sizes for achieving a specified power level.
### Usage Example for `mean_absolute_deviation`
```python
mad = mean_absolute_deviation([1, 2, 3, 4, 5])
print(f'Mean Absolute Deviation: {mad}')
```
### Usage Example for `calculate_chi_squared` and `calculate_expected`
```python
observed = np.array([[10, 10, 20], [20, 20, 10]])
expected = calculate_expected(observed)
chi_squared = calculate_chi_squared(observed, expected)
print(f'Chi-Squared Statistic: {chi_squared}')
```
### Usage Example for `bootstrap_chi_abs`
```python
bootstrap_results = bootstrap_chi_abs(observed_data, num_simulations=1000)
print('Bootstrap Chi Absolute Results:', bootstrap_results[:5])
```
### Usage Example for `calculate_relative_risk_two_treatments`
```python
relative_risk = calculate_relative_risk_two_treatments(observed_data, 0, 1, 2)
print(f'Relative Risk: {relative_risk}')
```
### Usage Example for `permute_correlation`
```python
x = np.random.rand(100)
y = 2 * x + np.random.normal(0, 1, 100)
simulated_correlations = permute_correlation(x, y, 10000)
print('Simulated Correlation Coefficients:', simulated_correlations[:5])
```
### Usage Example for `bootstrap_confidence_interval` and `plot_bootstrap_lines`
```python
x = np.random.rand(100)
y = 2 * x + np.random.normal(0, 1, 100)
confidence_interval = bootstrap_confidence_interval(x, y, 1000, 95, 'slope')
print(f'Confidence Interval for Slope: {confidence_interval}')
plot_bootstrap_lines(x, y)
```
### Usage Example for `power_analysis`
```python
group1 = np.random.normal(5, 2, 100)
group2 = np.random.normal(5.5, 2, 100)
sample_sizes, achieved_power = power_analysis(obs_diff=0.5, group1=group1, group2=group2, alpha=0.05, power_threshold=0.8)
print(f'Required Sample Sizes: {sample_sizes}, Achieved Power: {achieved_power}')
```
Additional Notes
This documentation includes examples and explanations for key functions. Users are encouraged to refer to the inline comments within the `resamp.py` script for more detailed information on specific parameters and functionality. For complex statistical analyses, users should ensure their input data is correctly formatted and understand the statistical principles underlying their analyses.
Contact
For further assistance or to report issues, please contact the author, Vishanth Hari Raj, or the supervisor, Jane.
ANNEXURE I
Function Documentation
1. read_data
The read_data function reads data from either a DataFrame or a file path. If the input is a DataFrame, it makes a copy. If it's a file path, it checks the file extension (.xls, .xlsx, .csv) and reads the data accordingly. It then applies pandas' to_numeric function to convert the data to numeric types, with non-numeric values coerced to NaN.
Usage Example:
data = read_data('path_to_file.csv') # For CSV files
data = read_data('path_to_file.xlsx') # For Excel files
data = read_data(dataframe) # For existing DataFrame
2. compare_dimensions
This function compares the dimensions of the observed and expected datasets. If the dimensions don't match, it raises a ValueError. This ensures that further statistical calculations are valid.
Usage Example:
compare_dimensions(observed_data, expected_data) # Checks if dimensions of observed and expected data match
3. calculate_chi_squared
Calculates the chi-squared statistic from observed and expected datasets. This is done by summing the squared differences between observed and expected values, divided by the expected values.
Usage Example:
chi_squared = calculate_chi_squared(observed_data, expected_data) # Returns the chi-squared statistic
4. calculate_expected
Computes expected frequencies for a contingency table based on the observed data. It uses row and column sums to calculate these frequencies, assuming independence between rows and columns.
Usage Example:
expected_data = calculate_expected(observed_data) # Returns the expected frequencies
5. calculate_chi_abs
This function calculates the chi absolute statistic for observed and expected data. If expected data is not provided, it calculates the expected data based on the observed data.
Usage Example:
chi_abs = calculate_chi_abs(observed_data, expected_data) # Calculates the chi absolute statistic
6. chi_abs_stat
A wrapper function for calculate_chi_abs. It allows the user to provide only observed data, with an option to provide expected data. It handles dimension comparison and error logging.
Usage Example:
chi_abs = chi_abs_stat(observed_data, expected_data) # Calculates chi absolute statistic with optional expected data
7. calculate_p_value
Calculates the p-value from a chi-squared statistic and degrees of freedom using the chi-squared survival function.
Usage Example:
p_value = calculate_p_value(chi_squared, dof) # Returns the p-value
8. chi_squared_stat
Reads observed and expected data, checks their dimensions, calculates the chi-squared statistic, and then computes the p-value.
Usage Example:
chi_squared_value = chi_squared_stat(observed_data, expected_data) # Calculates chi-squared statistic
9. p_value_stat
Reads observed and expected data, checks dimensions, calculates the chi-squared statistic, and then computes the p-value.
Usage Example:
p_value = p_value_stat(observed_data, expected_data) # Calculates the p-value for chi-squared statistic
10. convert_df_to_numpy
Converts DataFrame data to numpy arrays for use in functions, particularly for bootstrapping. It returns a tuple of numpy arrays for observed and expected data.
Usage Example:
observed_array, expected_array = convert_df_to_numpy(df_observed, df_expected) # Converts DataFrame to numpy arrays
11. bootstrap_chi_abs_distribution
Generates a bootstrap distribution of the chi absolute statistic for an n*n contingency table. Simulates new datasets and calculates the chi absolute for each, returning an array of these statistics.
Usage Example:
simulated_chi_abs = bootstrap_chi_abs_distribution(observed_data) # Returns array of simulated chi absolute statistics
12. calculate_p_value_bootstrap
Calculates the p-value for the chi absolute statistic using bootstrap methods. Compares the observed chi absolute statistic against the distribution of simulated chi absolute values to compute the p-value.
Usage Example:
p_value = calculate_p_value_bootstrap(observed_chi_abs, simulated_chi_abs) # Calculates p-value using bootstrap method
13. plot_chi_abs_distribution
Plots the distribution of simulated chi absolute values along with the observed chi absolute value. Shows the calculated p-value, providing a visual representation of the statistical analysis.
Usage Example:
plot_chi_abs_distribution(simulated_data, observed_data, p_value) # Plots distribution of chi absolute values
Extended:- Relative Risk Analysis Documentation
This documentation provides an overview and usage guide for a set of Python functions designed to calculate the relative risk between two treatments, resample data for statistical analysis, calculate confidence intervals, and plot the distribution of relative risks. These functions are intended for statistical analysis in lifesciences research or any field requiring comparative risk assessment.
1. calculate_relative_risk_two_treatments
Calculates the relative risk (RR) of an event occurring between two treatments. RR is a measure of the strength of association between an exposure and an outcome.
Parameters:
- observed_data: 2D array of observed data.
- event_row_index: Index of the row corresponding to the event.
- treatment1_index: Column index for the first treatment.
- treatment2_index: Column index for the second treatment.
Logic:
The function sums the total occurrences for each treatment and calculates the probability of the event for each treatment. The relative risk is then the ratio of these probabilities.
Usage Example:
relative_risk = calculate_relative_risk_two_treatments(observed_data, 0, 1, 2)
2. resample_and_calculate_rr
Performs resampling to calculate a distribution of relative risks through bootstrapping.
Parameters:
- observed_data: 2D array of observed data.
- event_row_index: Index of the event row.
- reference_treatment_index: Index for the reference treatment column (default is 0).
- num_simulations: Number of bootstrap simulations to perform.
Logic:
Generates simulated datasets by resampling from the observed data, maintaining the original proportions. Calculates the relative risk for each simulated dataset to create a distribution of relative risks.
Usage Example:
simulated_rr = resample_and_calculate_rr(observed_data, 0, num_simulations=10000)
3. calculate_confidence_interval
Calculates the 95% confidence interval for the relative risk from a distribution of simulated relative risks.
Parameters:
- simulated_rr (np.array): Array of simulated relative risks.
- percentile (float, optional): The percentile for the confidence interval, defaulting to 95%.
Logic:
The function calculates the lower and upper bounds of the confidence interval based on the specified percentile using numpy's percentile function. This provides an estimate of the interval within which the true relative risk is expected to lie with 95% certainty.
Usage Example:
lower_bound, upper_bound = calculate_confidence_interval(simulated_rr, 95)
4. calculate_probabilities_for_each_treatment
Calculates the probability of an event occurring for each treatment group.
Parameters:
- observed_data (np.array or pd.DataFrame): The observed data.
- event_row_index (int): The row index for the event.
Logic:
For each treatment group, the function calculates the total counts and then determines the probability of the event. These probabilities are stored in a dictionary, providing a quick reference for each treatment's event probability.
Usage Example:
probabilities = calculate_probabilities_for_each_treatment(observed_data, 0)
5. plot_relative_risk_distribution
Plots the distribution of simulated relative risks against the observed relative risk, including confidence intervals.
Parameters:
- simulated_rr (np.array): Array of simulated relative risks.
- observed_rr (float): The observed relative risk.
Logic:
This function visualizes the distribution of simulated relative risks, highlighting the observed relative risk and marking the confidence intervals. It uses seaborn and matplotlib for plotting.
Usage Example:
plot_relative_risk_distribution(simulated_rr, observed_rr)
APPENDIX - II
bootstrap_chi_abs_distribution :
The bootstrap_chi_abs_distribution function is designed for generating a bootstrap distribution of the chi absolute statistic for an n*n contingency table. This is crucial in statistical analysis, especially when assessing the significance of observed frequencies against expected frequencies in categorical data.
Parameters:
- observed_data: Observed frequencies in an n*n contingency table, either a numpy array or pandas DataFrame.
- num_simulations: Number of bootstrap samples to generate.
- with_replacement: Boolean indicating whether sampling should be with replacement.
Process:
1. Convert DataFrame to Numpy Array: If observed_data is a DataFrame, it's converted to a numpy array.
2. Determine Dimensions and Calculate Expected Frequencies: Calculates total rows, total columns, and expected frequencies.
3. Initialize Results Array: An array to store chi absolute statistics for each simulation.
4. Calculate Total Counts Per Column: Computes the total count of observations per column.
5. Create Pooled Data Array: Concatenates all categories for resampling.
6. Bootstrap Simulation Loop: Iterates num_simulations times, creating and analyzing simulated datasets.
7. Calculate Chi Absolute Statistic: Computes chi absolute statistic for each simulated dataset.
8. Return Results: Returns the array containing chi absolute statistics from all simulations.
Usage:
This function is used in statistical hypothesis testing to assess the significance of the observed data. By comparing the observed chi absolute statistic to a distribution generated through bootstrapping, one can infer the likelihood of observing such a statistic under the null hypothesis.
APPENDIX III
Raw data
{
"_id": null,
"home_page": "https://github.com/vishanth10/resamp.git",
"name": "resamp",
"maintainer": null,
"docs_url": null,
"requires_python": null,
"maintainer_email": null,
"keywords": "resampling techniques, resample, chi-abs, power analysis, relative risk, statistics, resampling chi-abs analysis",
"author": "Vishanth Hari Raj Balasubramanian",
"author_email": "rbvish1007@gmail.com",
"download_url": "https://files.pythonhosted.org/packages/e2/17/fbd42a8f2c78c7217f021db495ceb46db78fedce13fdda530b91d470e190/resamp-1.7.tar.gz",
"platform": null,
"description": "# Resampling Techniques Library for Statistical measure\n\nLibrary for resampling methods over Chi-Abs, correlation, power analysis, linear regression (beta value), relative risk -- LS 40 @UCLA Life science course\n\nRESAMP Library Documentation\n\n- Author: Vishanth Hari Raj\n- GitHub Repository: https://github.com/vishanth10/resamp\n\n\n`resamp.py` Version 1.6.8 Documentation\nOverview\n`resamp.py` is a comprehensive Python library designed for statistical analysis and resampling techniques. This document serves as a guide for utilizing the library's functions, including statistical tests, data analysis, and visualization tools.\nInstallation\nTo install `resamp.py`, run the following command in your terminal:\n```bash\npip install statistics_library\n```\nFunctions:\n\nMedian Absolute Deviation (MAD)\n`median_absolute_deviation(data)`\nCalculates the Median Absolute Deviation of a dataset.\n- **Parameters**:\n - `data` (_array-like or str_): A 1D array or list containing the dataset, or a filename pointing to a CSV or Excel file containing a single column of numbers.\n- **Returns**:\n - _float_: The Median Absolute Deviation of the dataset.\n- **Usage Example**:\n ```python\n mad = median_absolute_deviation([1, 2, 3, 4, 5])\n print(mad)\n ```\n**Chi-Squared Test**\n`calculate_chi_squared(observed, expected)`\nCalculates the chi-squared statistic given observed and expected frequencies.\n- **Parameters**:\n - `observed` (_DataFrame_): Observed frequencies.\n - `expected` (_DataFrame_): Expected frequencies.\n- **Returns**:\n - _float_: The chi-squared statistic.\nBootstrap Chi Absolute\n`bootstrap_chi_abs(observed_data, num_simulations=10000, with_replacement=True)`\nGenerates a bootstrap distribution of the chi absolute statistic for an n*n contingency table.\n- **Parameters**:\n - `observed_data` (_np.array or pd.DataFrame_): Contingency table with observed frequencies.\n - `num_simulations` (_int_): Number of bootstrap samples to generate.\n - `with_replacement` (_bool_): Indicates whether sampling should be with replacement.\n- **Returns**:\n - _np.array_: Bootstrap distribution of chi absolute values.\n \n**Relative Risk Analysis**\n`calculate_relative_risk_two_treatments(observed_data, event_row_index, treatment1_index, treatment2_index)`\nCalculates the relative risk of an event between two treatments.\n- **Parameters**:\n - `observed_data` (_np.array_): The observed data as a 2D array.\n - `event_row_index` (_int_): Row index for the event.\n - `treatment1_index` (_int_): Column index for the first treatment.\n - `treatment2_index` (_int_): Column index for the second treatment.\n- **Returns**:\n - _float_: The relative risk.\n - \n**Correlation Analysis**\n`permute_correlation(x, y, num_simulations=10000)`\nGenerates simulated correlation coefficients by permuting one variable.\n- **Parameters**:\n - `x` (_np.array_): Values of variable 1.\n - `y` (_np.array_): Values of variable 2.\n - `num_simulations` (_int_): Number of permutations.\n- **Returns**:\n - _np.array_: Simulated correlation coefficients.\n \n**Linear Regression Analysis**\n`bootstrap_confidence_interval(x, y, n_bootstrap=1000, confidence_level=99, return_type='both')`\nCalculates bootstrap confidence intervals for the slope and intercept of a linear regression model.\n- **Parameters**:\n - `x`, `y` (_np.array_): Predictor and response variables.\n - `n_bootstrap` (_int_): Number of bootstrap samples.\n - `confidence_level` (_float_): Confidence level for the interval.\n - `return_type` (_str_): Determines if the output is for 'slope', 'intercept', or 'both'.\n- **Returns**:\n - _dict_: Confidence intervals for the slope and/or intercept.\n\n**Power Analysis**\n`power_analysis(...)`\nPerforms a power analysis to determine the required sample size for achieving a specified power level.\n- **Parameters**:\n - `obs_diff`: the observed difference in medians or means between the two groups, depending on 'measure'.\n - `group1`: data for group 1.\n - `group2`: data for group 2.\n - `num_simulations`: number of simulations to perform.\n - `alpha`: significance level.\n - `power_threshold`: desired power level to achieve.\n - `factor_limit`: the maximum factor by which to increase the sample size.\n - `measure`: 'median' or 'mean', the statistical measure to use for comparison.\n - `verbose`: if True, print intermediate results.\n- **Returns**:\n - `required_sample_sizes`: a tuple of the required sample sizes for group 1 and group 2 to achieve the desired power.\n - `achieved_power`: the power that was achieved with the returned sample sizes.\n\n### Usage Example for `power_analysis`\n```python\ngroup1 = np.random.normal(5, 2, 100)\ngroup2 = np.random.normal(5.5, 2, 100)\nsample_sizes, achieved_power = power_analysis(obs_diff=0.5, group1=group1, group2=group2, alpha=0.05, power_threshold=0.8)\nprint(f'Required Sample Sizes: {sample_sizes}, Achieved Power: {achieved_power}')\n```\n\n**#Function Usage**\n\n```python\ndef mean_absolute_deviation(data):\n # Your implementation here\n```\nCalculates the mean absolute deviation of the provided dataset. Supports direct list input or file paths.\n```python\ndef calculate_chi_squared(observed, expected):\n # Your implementation here\n```\nCalculates the chi-squared statistic based on observed and expected frequencies.\n```python\ndef calculate_expected(observed):\n # Your implementation here\n```\nCalculates expected frequencies for a chi-squared test from observed data.\n```python\ndef bootstrap_chi_abs(observed_data, num_simulations=10000, with_replacement=True):\n # Your implementation here\n```\nGenerates bootstrap samples for chi absolute statistics and calculates the chi absolute statistic for each sample.\n```python\ndef calculate_relative_risk_two_treatments(observed_data, event_row_index, treatment1_index, treatment2_index):\n # Your implementation here\n```\nCalculates the relative risk between two treatments.\n```python\ndef permute_correlation(x, y, num_simulations=10000):\n # Your implementation here\n```\nPerforms permutation tests to calculate the null distribution of the correlation coefficient.\n```python\ndef bootstrap_confidence_interval(x, y, n_bootstrap=1000, confidence_level=99, return_type='both'):\n # Your implementation here\n```\nCalculates bootstrap confidence intervals for linear regression parameters.\n```python\ndef plot_bootstrap_lines(x, y, n_bootstrap=1000, original_slope=2, original_intercept=0):\n # Your implementation here\n```\nPlots regression lines for bootstrap samples and compares them with the original regression line.\n```python\ndef power_analysis(...):\n # Your implementation here\n```\nPerforms a power analysis to determine the required sample sizes for achieving a specified power level.\n### Usage Example for `mean_absolute_deviation`\n```python\nmad = mean_absolute_deviation([1, 2, 3, 4, 5])\nprint(f'Mean Absolute Deviation: {mad}')\n```\n### Usage Example for `calculate_chi_squared` and `calculate_expected`\n```python\nobserved = np.array([[10, 10, 20], [20, 20, 10]])\nexpected = calculate_expected(observed)\nchi_squared = calculate_chi_squared(observed, expected)\nprint(f'Chi-Squared Statistic: {chi_squared}')\n```\n### Usage Example for `bootstrap_chi_abs`\n```python\nbootstrap_results = bootstrap_chi_abs(observed_data, num_simulations=1000)\nprint('Bootstrap Chi Absolute Results:', bootstrap_results[:5])\n```\n### Usage Example for `calculate_relative_risk_two_treatments`\n```python\nrelative_risk = calculate_relative_risk_two_treatments(observed_data, 0, 1, 2)\nprint(f'Relative Risk: {relative_risk}')\n```\n### Usage Example for `permute_correlation`\n```python\nx = np.random.rand(100)\ny = 2 * x + np.random.normal(0, 1, 100)\nsimulated_correlations = permute_correlation(x, y, 10000)\nprint('Simulated Correlation Coefficients:', simulated_correlations[:5])\n```\n### Usage Example for `bootstrap_confidence_interval` and `plot_bootstrap_lines`\n```python\nx = np.random.rand(100)\ny = 2 * x + np.random.normal(0, 1, 100)\nconfidence_interval = bootstrap_confidence_interval(x, y, 1000, 95, 'slope')\nprint(f'Confidence Interval for Slope: {confidence_interval}')\nplot_bootstrap_lines(x, y)\n```\n### Usage Example for `power_analysis`\n```python\ngroup1 = np.random.normal(5, 2, 100)\ngroup2 = np.random.normal(5.5, 2, 100)\nsample_sizes, achieved_power = power_analysis(obs_diff=0.5, group1=group1, group2=group2, alpha=0.05, power_threshold=0.8)\nprint(f'Required Sample Sizes: {sample_sizes}, Achieved Power: {achieved_power}')\n```\n\nAdditional Notes\nThis documentation includes examples and explanations for key functions. Users are encouraged to refer to the inline comments within the `resamp.py` script for more detailed information on specific parameters and functionality. For complex statistical analyses, users should ensure their input data is correctly formatted and understand the statistical principles underlying their analyses.\nContact\nFor further assistance or to report issues, please contact the author, Vishanth Hari Raj, or the supervisor, Jane.\n\n\n\nANNEXURE I\n\n\n \nFunction Documentation\n1. read_data\nThe read_data function reads data from either a DataFrame or a file path. If the input is a DataFrame, it makes a copy. If it's a file path, it checks the file extension (.xls, .xlsx, .csv) and reads the data accordingly. It then applies pandas' to_numeric function to convert the data to numeric types, with non-numeric values coerced to NaN.\nUsage Example:\ndata = read_data('path_to_file.csv') # For CSV files\ndata = read_data('path_to_file.xlsx') # For Excel files\ndata = read_data(dataframe) # For existing DataFrame\n \n2. compare_dimensions\nThis function compares the dimensions of the observed and expected datasets. If the dimensions don't match, it raises a ValueError. This ensures that further statistical calculations are valid.\nUsage Example:\ncompare_dimensions(observed_data, expected_data) # Checks if dimensions of observed and expected data match\n \n3. calculate_chi_squared\nCalculates the chi-squared statistic from observed and expected datasets. This is done by summing the squared differences between observed and expected values, divided by the expected values.\nUsage Example:\nchi_squared = calculate_chi_squared(observed_data, expected_data) # Returns the chi-squared statistic\n \n4. calculate_expected\nComputes expected frequencies for a contingency table based on the observed data. It uses row and column sums to calculate these frequencies, assuming independence between rows and columns.\nUsage Example:\nexpected_data = calculate_expected(observed_data) # Returns the expected frequencies\n \n5. calculate_chi_abs\nThis function calculates the chi absolute statistic for observed and expected data. If expected data is not provided, it calculates the expected data based on the observed data.\nUsage Example:\nchi_abs = calculate_chi_abs(observed_data, expected_data) # Calculates the chi absolute statistic\n \n6. chi_abs_stat\nA wrapper function for calculate_chi_abs. It allows the user to provide only observed data, with an option to provide expected data. It handles dimension comparison and error logging.\nUsage Example:\nchi_abs = chi_abs_stat(observed_data, expected_data) # Calculates chi absolute statistic with optional expected data\n \n7. calculate_p_value\nCalculates the p-value from a chi-squared statistic and degrees of freedom using the chi-squared survival function.\nUsage Example:\np_value = calculate_p_value(chi_squared, dof) # Returns the p-value\n \n8. chi_squared_stat\nReads observed and expected data, checks their dimensions, calculates the chi-squared statistic, and then computes the p-value.\nUsage Example:\nchi_squared_value = chi_squared_stat(observed_data, expected_data) # Calculates chi-squared statistic\n \n9. p_value_stat\nReads observed and expected data, checks dimensions, calculates the chi-squared statistic, and then computes the p-value.\nUsage Example:\np_value = p_value_stat(observed_data, expected_data) # Calculates the p-value for chi-squared statistic\n \n10. convert_df_to_numpy\nConverts DataFrame data to numpy arrays for use in functions, particularly for bootstrapping. It returns a tuple of numpy arrays for observed and expected data.\nUsage Example:\nobserved_array, expected_array = convert_df_to_numpy(df_observed, df_expected) # Converts DataFrame to numpy arrays\n \n11. bootstrap_chi_abs_distribution\nGenerates a bootstrap distribution of the chi absolute statistic for an n*n contingency table. Simulates new datasets and calculates the chi absolute for each, returning an array of these statistics.\nUsage Example:\nsimulated_chi_abs = bootstrap_chi_abs_distribution(observed_data) # Returns array of simulated chi absolute statistics\n \n12. calculate_p_value_bootstrap\nCalculates the p-value for the chi absolute statistic using bootstrap methods. Compares the observed chi absolute statistic against the distribution of simulated chi absolute values to compute the p-value.\nUsage Example:\np_value = calculate_p_value_bootstrap(observed_chi_abs, simulated_chi_abs) # Calculates p-value using bootstrap method\n \n13. plot_chi_abs_distribution\nPlots the distribution of simulated chi absolute values along with the observed chi absolute value. Shows the calculated p-value, providing a visual representation of the statistical analysis.\nUsage Example:\nplot_chi_abs_distribution(simulated_data, observed_data, p_value) # Plots distribution of chi absolute values\n \n\n\nExtended:- Relative Risk Analysis Documentation\nThis documentation provides an overview and usage guide for a set of Python functions designed to calculate the relative risk between two treatments, resample data for statistical analysis, calculate confidence intervals, and plot the distribution of relative risks. These functions are intended for statistical analysis in lifesciences research or any field requiring comparative risk assessment.\n \n1. calculate_relative_risk_two_treatments\nCalculates the relative risk (RR) of an event occurring between two treatments. RR is a measure of the strength of association between an exposure and an outcome.\nParameters:\n- observed_data: 2D array of observed data.\n- event_row_index: Index of the row corresponding to the event.\n- treatment1_index: Column index for the first treatment.\n- treatment2_index: Column index for the second treatment.\nLogic:\nThe function sums the total occurrences for each treatment and calculates the probability of the event for each treatment. The relative risk is then the ratio of these probabilities.\nUsage Example:\nrelative_risk = calculate_relative_risk_two_treatments(observed_data, 0, 1, 2)\n \n2. resample_and_calculate_rr\nPerforms resampling to calculate a distribution of relative risks through bootstrapping.\nParameters:\n- observed_data: 2D array of observed data.\n- event_row_index: Index of the event row.\n- reference_treatment_index: Index for the reference treatment column (default is 0).\n- num_simulations: Number of bootstrap simulations to perform.\nLogic:\nGenerates simulated datasets by resampling from the observed data, maintaining the original proportions. Calculates the relative risk for each simulated dataset to create a distribution of relative risks.\nUsage Example:\nsimulated_rr = resample_and_calculate_rr(observed_data, 0, num_simulations=10000)\n \n3. calculate_confidence_interval\nCalculates the 95% confidence interval for the relative risk from a distribution of simulated relative risks.\nParameters:\n- simulated_rr (np.array): Array of simulated relative risks.\n- percentile (float, optional): The percentile for the confidence interval, defaulting to 95%.\nLogic:\nThe function calculates the lower and upper bounds of the confidence interval based on the specified percentile using numpy's percentile function. This provides an estimate of the interval within which the true relative risk is expected to lie with 95% certainty.\nUsage Example:\nlower_bound, upper_bound = calculate_confidence_interval(simulated_rr, 95)\n \n4. calculate_probabilities_for_each_treatment\nCalculates the probability of an event occurring for each treatment group.\nParameters:\n- observed_data (np.array or pd.DataFrame): The observed data.\n- event_row_index (int): The row index for the event.\nLogic:\nFor each treatment group, the function calculates the total counts and then determines the probability of the event. These probabilities are stored in a dictionary, providing a quick reference for each treatment's event probability.\nUsage Example:\nprobabilities = calculate_probabilities_for_each_treatment(observed_data, 0)\n \n5. plot_relative_risk_distribution\nPlots the distribution of simulated relative risks against the observed relative risk, including confidence intervals.\nParameters:\n- simulated_rr (np.array): Array of simulated relative risks.\n- observed_rr (float): The observed relative risk.\nLogic:\nThis function visualizes the distribution of simulated relative risks, highlighting the observed relative risk and marking the confidence intervals. It uses seaborn and matplotlib for plotting.\nUsage Example:\nplot_relative_risk_distribution(simulated_rr, observed_rr)\n\n\nAPPENDIX - II\n\nbootstrap_chi_abs_distribution :\nThe bootstrap_chi_abs_distribution function is designed for generating a bootstrap distribution of the chi absolute statistic for an n*n contingency table. This is crucial in statistical analysis, especially when assessing the significance of observed frequencies against expected frequencies in categorical data.\nParameters:\n- observed_data: Observed frequencies in an n*n contingency table, either a numpy array or pandas DataFrame.\n- num_simulations: Number of bootstrap samples to generate.\n- with_replacement: Boolean indicating whether sampling should be with replacement.\nProcess:\n1. Convert DataFrame to Numpy Array: If observed_data is a DataFrame, it's converted to a numpy array.\n2. Determine Dimensions and Calculate Expected Frequencies: Calculates total rows, total columns, and expected frequencies.\n3. Initialize Results Array: An array to store chi absolute statistics for each simulation.\n4. Calculate Total Counts Per Column: Computes the total count of observations per column.\n5. Create Pooled Data Array: Concatenates all categories for resampling.\n6. Bootstrap Simulation Loop: Iterates num_simulations times, creating and analyzing simulated datasets.\n7. Calculate Chi Absolute Statistic: Computes chi absolute statistic for each simulated dataset.\n8. Return Results: Returns the array containing chi absolute statistics from all simulations.\nUsage:\nThis function is used in statistical hypothesis testing to assess the significance of the observed data. By comparing the observed chi absolute statistic to a distribution generated through bootstrapping, one can infer the likelihood of observing such a statistic under the null hypothesis.\n\n\n\nAPPENDIX III\n \n\n",
"bugtrack_url": null,
"license": null,
"summary": "A custom statistics library of resampling technqiues for chi-abs, boostrapping analysis and other statistical functions.",
"version": "1.7",
"project_urls": {
"Homepage": "https://github.com/vishanth10/resamp.git"
},
"split_keywords": [
"resampling techniques",
" resample",
" chi-abs",
" power analysis",
" relative risk",
" statistics",
" resampling chi-abs analysis"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "5526a938291168a30e21f20a67529ae7b4c6d4b3ba4557d604c3b6ddd4656512",
"md5": "8113ae18ca7aa7be9641158e5b44a95c",
"sha256": "3555e800b929c31cbe39741c86e10ac3bdc1f1aaa2438be53d652c913e2acb5f"
},
"downloads": -1,
"filename": "resamp-1.7-py3-none-any.whl",
"has_sig": false,
"md5_digest": "8113ae18ca7aa7be9641158e5b44a95c",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 16911,
"upload_time": "2024-05-04T19:45:34",
"upload_time_iso_8601": "2024-05-04T19:45:34.958574Z",
"url": "https://files.pythonhosted.org/packages/55/26/a938291168a30e21f20a67529ae7b4c6d4b3ba4557d604c3b6ddd4656512/resamp-1.7-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "e217fbd42a8f2c78c7217f021db495ceb46db78fedce13fdda530b91d470e190",
"md5": "988d86d9faa52c43a2fb79f23cc5d173",
"sha256": "0351614132cb42ac48e4737e6ec6e9c3389eb73b209e4aa23e831de62a5ddf57"
},
"downloads": -1,
"filename": "resamp-1.7.tar.gz",
"has_sig": false,
"md5_digest": "988d86d9faa52c43a2fb79f23cc5d173",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 20991,
"upload_time": "2024-05-04T19:45:36",
"upload_time_iso_8601": "2024-05-04T19:45:36.756861Z",
"url": "https://files.pythonhosted.org/packages/e2/17/fbd42a8f2c78c7217f021db495ceb46db78fedce13fdda530b91d470e190/resamp-1.7.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-05-04 19:45:36",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "vishanth10",
"github_project": "resamp",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "resamp"
}
JSON
