AutoStatLib 0.1.6__py3-none-any.whl

AutoStatLib/AutoStatLib.py

@@ -0,0 +1,605 @@
+import numpy as np
+from statsmodels.stats.diagnostic import lilliefors
+from scipy.stats import ttest_rel, ttest_ind, ttest_1samp, wilcoxon, mannwhitneyu, f_oneway, kruskal, friedmanchisquare, shapiro, kstest, anderson, normaltest
+
+
+class __StatisticalTests():
+    '''
+        Statistical tests mixin
+    '''
+
+    def anova(self):
+        stat, p_value = f_oneway(*self.data)
+        self.tails = 2
+        # if self.tails == 1 and p_value > 0.5:
+        #     p_value /= 2
+        # if self.tails == 1:
+        #     p_value /= 2
+        self.test_name = 'ANOVA'
+        self.test_id = 'anova'
+        self.paired = False
+        self.test_stat = stat
+        self.p_value = p_value
+
+    def friedman_test(self):
+        stat, p_value = friedmanchisquare(*self.data)
+        self.tails = 2
+        self.test_name = 'Friedman test'
+        self.test_id = 'friedman'
+        self.paired = True
+        self.test_stat = stat
+        self.p_value = p_value
+
+    def kruskal_wallis_test(self):
+        stat, p_value = kruskal(*self.data)
+        self.test_name = 'Kruskal-Wallis test'
+        self.test_id = 'kruskal_wallis'
+        self.paired = False
+        self.test_stat = stat
+        self.p_value = p_value
+
+    def mann_whitney_u_test(self):
+        stat, p_value = mannwhitneyu(
+            self.data[0], self.data[1], alternative='two-sided')
+        if self.tails == 1:
+            p_value /= 2
+        # alternative method of one-tailed calculation
+        # gives the same result:
+        # stat, p_value = mannwhitneyu(
+        #     self.data[0], self.data[1], alternative='two-sided' if self.tails == 2 else 'less')
+        # if self.tails == 1 and p_value > 0.5:
+        #     p_value = 1-p_value
+
+        self.test_name = 'Mann-Whitney U test'
+        self.test_id = 'mann_whitney'
+        self.paired = False
+        self.test_stat = stat
+        self.p_value = p_value
+
+    def t_test_independend(self):
+        t_stat, t_p_value = ttest_ind(
+            self.data[0], self.data[1])
+        if self.tails == 1:
+            t_p_value /= 2
+        self.test_name = 't-test for independend samples'
+        self.test_id = 't_test_independend'
+        self.paired = False
+        self.test_stat = t_stat
+        self.p_value = t_p_value
+
+    def t_test_paired(self):
+        t_stat, t_p_value = ttest_rel(
+            self.data[0], self.data[1])
+        if self.tails == 1:
+            t_p_value /= 2
+        self.test_name = 't-test for paired samples'
+        self.test_id = 't_test_paired'
+        self.paired = True
+        self.test_stat = t_stat
+        self.p_value = t_p_value
+
+    def t_test_single_sample(self):
+        if self.popmean == None:
+            self.popmean = 0
+            self.AddWarning('no_pop_mean_set')
+        t_stat, t_p_value = ttest_1samp(self.data[0], self.popmean)
+        if self.tails == 1:
+            t_p_value /= 2
+        self.test_name = 'Single-sample t-test'
+        self.test_id = 't_test_single_sample'
+        self.paired = False
+        self.test_stat = t_stat
+        self.p_value = t_p_value
+
+    def wilcoxon_single_sample(self):
+        if self.popmean == None:
+            self.popmean = 0
+            self.AddWarning('no_pop_mean_set')
+        data = [i - self.popmean for i in self.data[0]]
+        w_stat, w_p_value = wilcoxon(data)
+        if self.tails == 1:
+            p_value /= 2
+        self.test_name = 'Wilcoxon signed-rank test for single sample'
+        self.test_id = 'wilcoxon_single_sample'
+        self.paired = False
+        self.test_stat = w_stat
+        self.p_value = w_p_value
+
+    def wilcoxon(self):
+        stat, p_value = wilcoxon(self.data[0], self.data[1])
+        if self.tails == 1:
+            p_value /= 2
+        self.test_name = 'Wilcoxon signed-rank test'
+        self.test_id = 'wilcoxon'
+        self.paired = True
+        self.test_stat = stat
+        self.p_value = p_value
+
+
+class __NormalityTests():
+    '''
+        Normality tests mixin
+
+        see the article about minimum sample size for tests:
+        Power comparisons of Shapiro-Wilk, Kolmogorov-Smirnov,
+        Lilliefors and Anderson-Darling tests, Nornadiah Mohd Razali1, Yap Bee Wah1
+    '''
+
+    def check_normality(self, data):
+        sw = None
+        lf = None
+        ad = None
+        ap = None
+        n = len(data)
+
+        # Shapiro-Wilk test
+        sw_stat, sw_p_value = shapiro(data)
+        if sw_p_value > 0.05:
+            sw = True
+        else:
+            sw = False
+
+        # Lilliefors test
+        lf_stat, lf_p_value = lilliefors(
+            data, dist='norm')
+        if lf_p_value > 0.05:
+            lf = True
+        else:
+            lf = False
+
+        # Anderson-Darling test
+        ad_stat, ad_p_value = self.anderson_get_p(
+            data, dist='norm')
+        if ad_p_value > 0.05 and n >= 20:
+            ad = True
+        elif ad_p_value <= 0.05 and n >= 20:
+            ad = False
+
+        # D'Agostino-Pearson test
+        ap_stat, ap_p_value = normaltest(data)
+        # test result is skewed if n<20
+        if ap_p_value > 0.05 and n >= 20:
+            ap = True
+        elif ap_p_value <= 0.05 and n >= 20:
+            ap = False
+
+        # print(ap_p_value, ad_p_value, sw_p_value, lf_p_value)
+
+        return (sw, lf, ad, ap)
+
+    def anderson_get_p(self, data, dist='norm'):
+        '''
+            calculating p-value for Anderson-Darling test using the method described here:            
+            Computation of Probability Associated with Anderson-Darling Statistic
+            Lorentz Jantschi and Sorana D. Bolboaca, 2018 - Mathematics
+
+        '''
+        e = 2.718281828459045
+        n = len(data)
+
+        ad, critical_values, significance_levels = anderson(
+            data, dist=dist)
+
+        # adjust ad_stat for small sample sizes:
+        s = ad*(1 + 0.75/n + 2.25/(n**2))
+
+        if s >= 0.6:
+            p = e**(1.2937 - 5.709*s + 0.0186*s**2)
+        elif s > 0.34:
+            p = e**(0.9177 - 4.279*s - 1.38*s**2)
+        elif s > 0.2:
+            p = 1 - e**(-8.318 + 42.796*s - 59.938*s**2)
+        elif s <= 0.2:
+            p = 1 - e**(-13.436 + 101.14*s - 223.73*s**2)
+        else:
+            p = None
+
+        return ad, p
+
+
+class __TextFormatting():
+    '''
+        Text formatting mixin
+    '''
+
+    def autospace(self, elements_list, space, delimiter=' ') -> str:
+        output = ''
+        for i, element in enumerate(elements_list):
+            if i == len(elements_list):
+                output += element
+            else:
+                output += element + (space-len(element))*delimiter
+        return output
+
+    def print_groups(self, space=24, max_length=15):
+        self.log('')
+        # Get the number of groups (rows) and the maximum length of rows
+        data = self.data
+        num_groups = len(data)
+        group_longest = max(len(row) for row in data)
+
+        # Print the header
+        header = [f'Group {i+1}' for i in range(num_groups)]
+        line = [''*7]
+        self.log(self.autospace(header, space))
+        self.log(self.autospace(line, space))
+
+        # Print each column with a placeholder if longer than max_length
+        for i in range(group_longest):
+            row_values = []
+            all_values_empty = True
+            for row in data:
+                if len(row) > max_length:
+                    if i < max_length:
+                        row_values.append(str(row[i]))
+                        all_values_empty = False
+                    elif i == max_length:
+                        row_values.append(f'[{len(row) - max_length} more]')
+                        all_values_empty = False
+                    else:
+                        continue
+                else:
+                    if i < len(row):
+                        row_values.append(str(row[i]))
+                        all_values_empty = False
+                    else:
+                        row_values.append('')
+            if all_values_empty:
+                break
+            self.log(self.autospace(row_values, space))
+
+    def make_stars(self) -> int:
+        p = self.p_value.item()
+        if p is not None:
+            if p < 0.0001:
+                return 4
+            if p < 0.001:
+                return 3
+            elif p < 0.01:
+                return 2
+            elif p < 0.05:
+                return 1
+            else:
+                return 0
+        return 0
+
+    def make_p_value_printed(self) -> str:
+        p = self.p_value.item()
+        if p is not None:
+            if p > 0.99:
+                return 'p>0.99'
+            elif p >= 0.01:
+                return f'p={p:.2g}'
+            elif p >= 0.001:
+                return f'p={p:.2g}'
+            elif p >= 0.0001:
+                return f'p={p:.1g}'
+            elif p < 0.0001:
+                return 'p<0.0001'
+            else:
+                return 'N/A'
+        return 'N/A'
+
+    def print_results(self):
+        self.log('\n\nResults: \n')
+        for i in self.results:
+            shift = 27 - len(i)
+            if i == 'Warnings':
+                self.log(i, ':', ' ' * shift, len(self.results[i]))
+            elif i == 'Samples':
+                pass
+            else:
+                self.log(i, ':', ' ' * shift, self.results[i])
+
+    def create_results_dict(self) -> dict:
+
+        self.stars_int = self.make_stars()
+        self.stars_str = '*' * self.stars_int if self.stars_int else 'ns'
+
+        return {
+            'p-value': self.make_p_value_printed(),
+            'Significance(p<0.05)':  True if self.p_value.item() < 0.05 else False,
+            'Stars_Printed': self.stars_str,
+            'Test_Name': self.test_name,
+            'Groups_Compared': self.n_groups,
+            'Population_Mean': self.popmean if self.n_groups == 1 else 'N/A',
+            'Data_Normaly_Distributed': self.parametric,
+            'Parametric_Test_Applied': True if self.test_id in self.test_ids_parametric else False,
+            'Paired_Test_Applied': self.paired,
+            'Tails': self.tails,
+            'p-value_exact': self.p_value.item(),
+            'Stars':  self.stars_int,
+            # 'Stat_Value': self.test_stat.item(),
+            'Warnings': self.warnings,
+            'Groups_N': [len(self.data[i]) for i in range(len(self.data))],
+            'Groups_Median': [np.median(self.data[i]).item() for i in range(len(self.data))],
+            'Groups_Mean': [np.mean(self.data[i]).item() for i in range(len(self.data))],
+            'Groups_SD': [np.std(self.data[i]).item() for i in range(len(self.data))],
+            'Groups_SE': [np.std(self.data[i]).item() / np.sqrt(len(self.data)).item() for i in range(len(self.data))],
+            # actually returns list of lists of numpy dtypes of float64, next make it return regular floats:
+            'Samples': self.data,
+        }
+
+    def log(self, *args, **kwargs):
+        message = ' '.join(map(str, args))
+        # print(message, **kwargs)
+        self.summary += '\n' + message
+
+    def AddWarning(self, warning_id):
+        message = self.warning_ids_all[warning_id]
+        self.log(message)
+        self.warnings.append(message)
+
+
+class __InputFormatting():
+    def floatify_recursive(self, data):
+        if isinstance(data, list):
+            # Recursively process sublists and filter out None values
+            processed_list = [self.floatify_recursive(item) for item in data]
+            return [item for item in processed_list if item is not None]
+        else:
+            try:
+                # Try to convert the item to float
+                return np.float64(data)
+            except (ValueError, TypeError):
+                # If conversion fails, replace with None
+                self.warning_flag_non_numeric_data = True
+                return None
+
+
+class StatisticalAnalysis(__StatisticalTests, __NormalityTests, __TextFormatting, __InputFormatting):
+    '''
+        The main class
+        *documentation placeholder*
+
+    '''
+
+    def __init__(self,
+                 groups_list,
+                 paired=False,
+                 tails=2,
+                 popmean=None,
+                 verbose=True):
+        self.results = None
+        self.error = False
+        self.groups_list = groups_list
+        self.paired = paired
+        self.tails = tails
+        self.popmean = popmean
+        self.verbose = verbose
+        self.n_groups = len(self.groups_list)
+        self.warning_flag_non_numeric_data = False
+        self.summary = ''
+        self.test_ids_parametric = ['anova',
+                                    't_test_independend',
+                                    't_test_paired',
+                                    't_test_single_sample',]
+        self.test_ids_all = [  # in aplhabetical order
+            'anova',
+            'friedman',
+            'kruskal_wallis',
+            'mann_whitney',
+            't_test_independend',
+            't_test_paired',
+            't_test_single_sample',
+            'wilcoxon',
+            'wilcoxon_single_sample',
+        ]
+        self.warning_ids_all = {
+            'not-numeric':                     '\nWarning: Non-numeric data was found in input and ignored.\n         Make sure the input data is correct to get the correct results\n',
+            'param_test_with_non-normal_data': '\nWarning: Parametric test was manualy chosen for Not-Normaly distributed data.\n         The results might be skewed. \n         Please, run non-parametric test or preform automatic test selection.\n',
+            'non-param_test_with_normal_data': '\nWarning: Non-Parametric test was manualy chosen for Normaly distributed data.\n         The results might be skewed. \n         Please, run parametric test or preform automatic test selection.\n',
+            'no_pop_mean_set':                 '\nWarning: No Population Mean was set up for single-sample test, used default 0 value.\n         The results might be skewed. \n         Please, set the Population Mean and run the test again.\n',
+        }
+
+    def __run_test(self, test='auto'):
+
+        # reset values from previous tests
+        self.results = None
+        self.error = False
+        self.warnings = []
+        self.normals = []
+        self.test_name = None
+        self.test_id = None
+        self.test_stat = None
+        self.p_value = None
+
+        self.log('\n' + '-'*67)
+        self.log('Statistical analysis initiated for data in {} groups\n'.format(
+            len(self.groups_list)))
+
+        # adjusting input data type
+        self.data = self.floatify_recursive(self.groups_list)
+        if self.warning_flag_non_numeric_data:
+            self.AddWarning('not-numeric')
+
+        # delete the empty cols from input
+        self.data = [col for col in self.data if any(
+            x is not None for x in col)]
+
+        # User input assertion block
+        try:
+            assert self.tails in [1, 2], 'Tails parameter can be 1 or 2 only'
+            assert test in self.test_ids_all or test == 'auto', 'Wrong test id choosen, ensure you called correct function'
+            assert not (self.n_groups > 1
+                        and (test == 't_test_single_sample'
+                             or test == 'wilcoxon_single_sample')), 'Only one group of data must be given for single-group tests'
+            assert all(len(
+                group) > 2 for group in self.data), 'Each group must contain at least three values'
+            assert not (self.paired == True and not all(len(lst) == len(
+                self.data[0]) for lst in self.data)), 'Paired groups must be the same length'
+            assert not (test == 'friedman' and not all(len(lst) == len(
+                self.data[0]) for lst in self.data)), 'Paired groups must be the same length for Friedman Chi Square test'
+            assert not (test == 't_test_paired' and not all(len(lst) == len(
+                self.data[0]) for lst in self.data)), 'Paired groups must be the same length for Paired t-test'
+            assert not (test == 'wilcoxon' and not all(len(lst) == len(
+                self.data[0]) for lst in self.data)), 'Paired groups must be the same length for Wilcoxon signed-rank test'
+            assert not (test == 'friedman' and self.n_groups <
+                        3), 'At least three groups of data must be given for 3-groups tests'
+            assert not ((test == 'anova'
+                         or test == 'kruskal_wallis') and self.n_groups < 2), 'At least two groups of data must be given for ANOVA or Kruskal Wallis tests'
+            assert not ((test == 'wilcoxon'
+                         or test == 't_test_independend'
+                         or test == 't_test_paired'
+                         or test == 'mann_whitney')
+                        and self.n_groups != 2), 'Only two groups of data must be given for 2-groups tests'
+        except AssertionError as error:
+            self.log('\nTest  :', test)
+            self.log('Error :', error)
+            self.log('-'*67 + '\n')
+            self.error = True
+            print(self.summary)
+            return
+
+        # Print the data
+        self.print_groups()
+
+        # Normality tests
+        self.log(
+            '\n\nThe group is assumed to be normally distributed if at least one')
+        self.log(
+            'normality test result is positive. Normality checked by tests:')
+        self.log('Shapiro-Wilk, Lilliefors, Anderson-Darling, D\'Agostino-Pearson')
+        self.log(
+            '[+] -positive, [-] -negative, [ ] -too small group for the test\n')
+        self.log('        Test   :   SW  LF  AD  AP  ')
+        for i, data in enumerate(self.data):
+            poll = self.check_normality(data)
+            isnormal = any(poll)
+            poll_print = tuple(
+                '+' if x is True else '-' if x is False else ' ' if x is None else 'e' for x in poll)
+            self.normals.append(isnormal)
+            self.log(
+                f'        Group {i+1}:    {poll_print[0]}   {poll_print[1]}   {poll_print[2]}   {poll_print[3]}   so disrtibution seems {"normal" if isnormal else "not normal"}')
+        self.parametric = all(self.normals)
+
+        # print test choosen
+        self.log('\n\nInput:\n')
+        self.log('Data Normaly Distributed:     ', self.parametric)
+        self.log('Paired Groups:                ', self.paired)
+        self.log('Groups:                       ', self.n_groups)
+        self.log('Test chosen by user:          ', test)
+
+        # Wrong test Warnings
+        if not test == 'auto' and not self.parametric and test in self.test_ids_parametric:
+            self.AddWarning('param_test_with_non-normal_data')
+        if not test == 'auto' and self.parametric and not test in self.test_ids_parametric:
+            self.AddWarning('non-param_test_with_normal_data')
+
+        if test == 'anova':
+            self.anova()
+        elif test == 'friedman':
+            self.friedman_test()
+        elif test == 'kruskal_wallis':
+            self.kruskal_wallis_test()
+        elif test == 'mann_whitney':
+            self.mann_whitney_u_test()
+        elif test == 't_test_independend':
+            self.t_test_independend()
+        elif test == 't_test_paired':
+            self.t_test_paired()
+        elif test == 't_test_single_sample':
+            self.t_test_single_sample()
+        elif test == 'wilcoxon':
+            self.wilcoxon()
+        elif test == 'wilcoxon_single_sample':
+            self.wilcoxon_single_sample()
+        else:
+            self.log('Automatic test selection preformed.')
+            self.__auto()
+
+        # print the results
+        self.results = self.create_results_dict()
+        self.print_results()
+        self.log(
+            '\n\nResults above are accessible as a dictionary via GetResult() method')
+        self.log('-'*67 + '\n')
+
+        # print the results to console:
+        if self.verbose == True:
+            print(self.summary)
+
+    def __auto(self):
+
+        if self.n_groups == 2:
+            if self.paired:
+                if self.parametric:
+                    return self.t_test_paired()
+                else:
+                    return self.wilcoxon()
+            else:
+                if self.parametric:
+                    return self.t_test_independend()
+                else:
+                    return self.mann_whitney_u_test()
+        elif self.n_groups == 1:
+            if self.parametric:
+                return self.t_test_single_sample()
+            else:
+                return self.wilcoxon_single_sample()
+        else:
+            if self.paired:
+                return self.friedman_test()
+            else:
+                if self.parametric:
+                    return self.anova()
+                else:
+                    return self.kruskal_wallis_test()
+
+    # public methods:
+    def RunAuto(self):
+        self.__run_test(test='auto')
+
+    def RunManual(self, test):
+        self.__run_test(test)
+
+    def RunAnova(self):
+        self.__run_test(test='anova')
+
+    def RunFriedman(self):
+        self.__run_test(test='friedman')
+
+    def RunKruskalWallis(self):
+        self.__run_test(test='kruskal_wallis')
+
+    def RunMannWhitney(self):
+        self.__run_test(test='mann_whitney')
+
+    def RunTtest(self):
+        self.__run_test(test='t_test_independend')
+
+    def RunTtestPaired(self):
+        self.__run_test(test='t_test_paired')
+
+    def RunTtestSingleSample(self):
+        self.__run_test(test='t_test_single_sample')
+
+    def RunWilcoxonSingleSample(self):
+        self.__run_test(test='wilcoxon_single_sample')
+
+    def RunWilcoxon(self):
+        self.__run_test(test='wilcoxon')
+
+    def GetResult(self):
+        if not self.results and not self.error:
+            self.__run_test(test='auto')
+            return self.results
+        if not self.results and self.error:
+            print('Error occured, no results to output')
+            return {}
+        else:
+            return self.results
+
+    def GetSummary(self):
+        if not self.results and not self.error:
+            self.__run_test(test='auto')
+            return self.summary
+        else:
+            return self.summary
+
+    def PrintSummary(self):
+        print(self.summary)
+
+
+if __name__ == '__main__':
+    print('\nThis package works as an imported module only\n')

AutoStatLib/__init__.py

@@ -0,0 +1,2 @@
+from .AutoStatLib import StatisticalAnalysis
+from ._version import __version__

AutoStatLib/__main__.py

@@ -0,0 +1,6 @@
+#!/usr/bin/env python
+from .AutoStatLib import StatisticalAnalysis
+from ._version import __version__
+
+if __name__ == '__main__':
+    print('\nThis package works as an imported module only\n')

AutoStatLib/_version.py

@@ -0,0 +1,2 @@
+# AutoStatLib package version:
+__version__ = "0.1.6"