def get_pass_rate(cut,
f_df,
c_df,
round_down=True,
grading_processed=True,
graded=None):
if graded is None:
if grading_processed:
graded = hr.grade_examination(f_df,
c_df,
correct=1,
incorrect=0,
only_operational=True,
grading_processed=grading_processed)
else:
graded = f_df
graded = graded.replace('1', float(1.0))
graded = graded.replace('2', float(0.0))
items = f_df.shape[1] - 1
required_to_pass = int(cut * items)
if not round_down:
required_to_pass += 1
graded['SCORE'] = graded.sum(axis=1)
graded['PASS'] = 0
graded.loc[graded['SCORE'] > required_to_pass, 'PASS'] = 1
max_score = graded['SCORE'].max()
high = graded.sort_values(by='SCORE')
return graded['PASS'].sum() / graded.shape[0]
def get_classical_stats(project):
forms = hfh.get_all_file_names_in_folder(project + '/forms/archival_forms')
ret = hfh.pd.DataFrame([])
for form in forms:
name = hfh.get_stem(form)[:-2]
processed_path = project + '/processed_data/'
f_path = processed_path + name + '_f.csv'
c_path = processed_path + name + '_c.csv'
if hfh.os.path.isfile(f_path) is False or hfh.os.path.isfile(
c_path) is False:
print("missing _f or _c for form " + form)
return False
f_df = hfh.get_df(f_path, header=None)
c_df = hfh.get_df(c_path, header=None)
#f_df.index = f_df.iloc[:,0]
#f_df = f_df.drop(columns = 'ID')
graded = hr.grade_examination(f_df,
c_df,
correct=1,
incorrect=0,
only_operational=True,
grading_processed=True)
classical_stats = hr.create_classical_stats_from_graded(
graded, form_name=hfh.get_stem(form)[:-2] + '_')
ret = hfh.pd.concat([ret, classical_stats], axis=1)
return ret
def create_classical_stats(project_path):
# creates a report per response string file
f_files = hfh.get_all_files(project_path + '/processed_data',
target_string='_f.csv')
c_files = hfh.get_all_files(project_path + '/processed_data',
target_string='_c.csv')
assert len(
c_files) == 1, "there should be exactly 1 _c file in processed data."
c_file = c_files[0]
c_df = hfh.get_df(c_file, header=0)
stats = []
for file in f_files:
f_df = hfh.get_df(file)
graded = hr.grade_examination(f_df,
c_df,
correct=1,
incorrect=0,
only_operational=False,
grading_processed=True)
form = hfh.get_stem(file)
_i = form.find(' ')
name = form[:_i]
classical_stats = hr.create_classical_stats_from_graded(graded,
form_name=name)
stats.append(classical_stats)
cumulative_df = hfh.pd.concat(stats, axis=1)
create_flagged(cumulative_df, project_path)
def update_formatted(combined_c_df, new_f_df, new_c_df, old_f_df, old_c_df):
#loop through control_df for each AccNum
graded = hr.grade_examination(new_f_df, new_c_df)
new = hfh.pd.DataFrame([], columns=combined_c_df.index)
for item in graded:
if item in new.columns:
new[item] = graded[item]
ret = hfh.pd.concat([old_f_df, new])
return ret
def create_classical_stats(self, form):
project_path = form
# creates a report per response string file
f_files = hfh.get_all_files(project_path + '/processed_data',
target_string='_f.csv')
c_files = hfh.get_all_files(project_path + '/processed_data',
target_string='_c.csv')
assert len(
c_files
) == 1, "there should be exactly 1 _c file in processed data."
c_file = c_files[0]
c_df = hfh.get_df(c_file, header=0)
stats = []
# creating classical aggregate report
for file in f_files:
if file.find("CUMULATIVE") == -1:
f_df = hfh.get_df(file)
graded = hr.grade_examination(f_df,
c_df,
correct=1,
incorrect=0,
only_operational=False)
form = hfh.get_stem(file)
_i = form.find(' ')
name = form[:_i]
classical_stats = hr.create_classical_stats_from_graded(
graded, form_name=name)
stats.append(classical_stats)
cumulative_df = hfh.pd.concat(stats, axis=1)
self.create_flagged(cumulative_df, project_path)
# creating item level classical report
item_level_classical_stats = []
cumulative_classical_stats = None
for file in f_files:
form = hfh.get_stem(file)
_i = form.find(' ')
df = hp.create_upload_from_processed(c_file, file)
new_cols = []
for col in df.columns:
new_cols.append(col + '_' + name)
df.columns = new_cols
item_level_classical_stats.append(df)
if file.find("CUMULATIVE") > -1:
df.to_csv(project_path + '/reports/' + name +
'_CUMULATIVE_ITEM_STATS.csv')
if len(item_level_classical_stats) > 0:
item_level = hfh.pd.concat(item_level_classical_stats, axis=1)
item_level.to_csv(project_path + '/reports/ugly/' + name +
'COMPLETE_ITEM_STATS_.csv')
def create_passing_comparison(project, cut_theta, form, angoff=None):
if not angoff:
angoff_cut_score = int(hfh.get_stem(form[form.find('_L') + 1:]))
else:
angoff_cut_score = angoff
if angoff is None:
name = hfh.get_stem(form)[:-4]
else:
name = hfh.get_stem(form)[:-2]
#todo make a canonical path for processed prior to drift removal
processed_path = project + '/backup_processed_data/'
f_path = processed_path + name + '_f.csv'
c_path = processed_path + name + '_c.csv'
if hfh.os.path.isfile(f_path) is False or hfh.os.path.isfile(
c_path) is False:
print("missing _f or _c for form " + form)
return False
f_df = hfh.get_df(f_path, header=None)
c_df = hfh.get_df(c_path, header=None)
#f_df.index = f_df.iloc[:,0]
#f_df = f_df.drop(columns = 'ID')
graded = hr.grade_examination(f_df,
c_df,
correct=1,
incorrect=0,
only_operational=True,
grading_processed=True)
ret = pd.DataFrame([])
ret['SCORE'] = graded.sum(axis=1)
p_item = graded.mean()
operational_items = graded.shape[1]
ret['PERCENT_SCORE'] = ret['SCORE'] / operational_items
cut_p = hp.create_passing(project, form, cut_theta, verbose=False)
if cut_p >= 1: cut_p /= 100
ret['ANGOFF_CUT_P'] = angoff_cut_score / operational_items
rasch_cut_score = int(cut_p * operational_items)
ret['ANGOFF_PASS'] = 0
ret.loc[ret['SCORE'] >= angoff_cut_score, 'ANGOFF_PASS'] = 1
ret['RASCH_CUT_P_' + str(cut_theta)] = cut_p
ret['RASCH_PASS_' + str(cut_theta)] = 0
ret.loc[ret['SCORE'] >= rasch_cut_score,
'RASCH_PASS_' + str(cut_theta)] = 1
ret = ret.mean()
ret.loc['FORM'] = hfh.get_stem(f_path)[:-2]
ret.loc['RASCH_CUT_SCORE_' + str(cut_theta)] = rasch_cut_score
ret.loc['ANGOFF_CUT_SCORE'] = angoff_cut_score
return ret
def get_angoff_passing_report(project, form, angoff=None, strict=True):
if angoff is None:
form = hfh.get_stem(form)
u = form.rfind('_')
name = hfh.get_stem(form)[:u]
angoff = int(form[u + 1:])
else:
name = hfh.get_stem(form)[:-2]
processed_path = project + '/backup_processed_data/'
f_path = processed_path + name + '_f.csv'
c_path = processed_path + name + '_c.csv'
if hfh.os.path.isfile(f_path) is False or hfh.os.path.isfile(
c_path) is False:
print("missing _f or _c for form " + form)
return False
f_df = hfh.get_df(f_path, header=None)
if strict:
c_df = hfh.get_df(c_path, header=0)
else:
c_df = hfh.get_df(c_path, header=None)
graded = hr.grade_examination(f_df,
c_df,
correct=1,
incorrect=0,
only_operational=True,
grading_processed=True)
ret = pd.DataFrame([])
ret['SCORE'] = graded.sum(axis=1)
p_item = graded.mean()
operational_items = graded.shape[1]
ret['PERCENT_SCORE'] = ret['SCORE'] / operational_items
#cut_p = hp.create_passing(project,form,cut_theta, verbose=False)
#if cut_p >=1: cut_p/=100
ret['ANGOFF_CUT_SCORE'] = angoff
ret['ANGOFF_CUT_P'] = angoff / operational_items
ret['ANGOFF_PASS_RATE'] = 0
ret.loc[ret['SCORE'] >= angoff, 'ANGOFF_PASS_RATE'] = 1
ret = ret.mean()
form_S = hfh.pd.Series([name], index=['FORM'])
ret = form_S.append(ret)
return ret, graded
def create_form_report(self, form, f_path, c_path):
# passing is defined as a number after the last _
project_path = form
date_start = hfh.get_stem(f_path).find('_')
date_finish = hfh.get_stem(f_path).rfind('_')
cumulative = False
if f_path.find("CUMULATIVE") > 0:
cumulative = True
ret = []
if date_start > 0 and date_finish > 0 or cumulative:
date = '_CUMULATIVE_'
if not cumulative:
date = hfh.get_stem(f_path)[date_start + 1:date_finish]
ret.append(["DATE", date])
bank_file = hfh.get_single_file(project_path, target_string='xlsx')
passing_score = bank_file[bank_file.rfind('_') +
1:bank_file.rfind('.')]
assert int(
passing_score), 'PASSING SCORE IS DEFINED AS NAME_SCORE.xlsx'
passing_score = int(passing_score)
f_df = hfh.get_df(f_path)
c_df = hfh.get_df(c_path, header=0)
graded_df = hr.grade_examination(f_df,
c_df,
grading_processed=True,
incorrect=0,
only_operational=True)
n = graded_df.shape[0]
ret.append(['N', n])
graded_df['SCORE'] = graded_df.sum(axis=1)
graded_df['PASS'] = graded_df['SCORE'] >= passing_score
proportion_of_candidates_who_pass = graded_df['PASS'].mean()
ret.append([
"PASS_P",
hfh.c_round(proportion_of_candidates_who_pass,
as_string=True,
as_percentage=True)
])
graded_df['MARGINAL'] = abs(passing_score - graded_df['SCORE']) < 2
marginal = sum(graded_df['MARGINAL']) / graded_df.shape[0]
if marginal > 0:
marginal = hfh.c_round(marginal,
as_string=True,
as_percentage=True)
else:
marginal = 0
ret.append(['BOARDERLINE', marginal])
average_pbis = graded_df.corr()['SCORE'].mean()
ret.append(['PBIS', hfh.c_round(average_pbis)])
average_score = graded_df['SCORE'].mean()
if average_score > 0:
average_score = hfh.c_round(average_score)
else:
average_score = 0
ret.append(['AVERAGE_SCORE', average_score])
top_10 = graded_df['SCORE'].quantile([.9])
top_10 = int(top_10.values[0])
ret.append(['TOP10', top_10])
bottom_10 = int(graded_df['SCORE'].quantile([.1]).values[0])
ret.append(['BOTTOM10', bottom_10])
SEM = graded_df['SCORE'].std() / pow(graded_df.shape[0], .5)
ret.append(['SEM', hfh.c_round(SEM)])
alpha = self.get_alpha(graded_df)
ret.append(['ALPHA', hfh.c_round(alpha)])
min = graded_df['SCORE'].min()
ret.append(['MIN_S', min])
max = graded_df['SCORE'].max()
ret.append(['MAX_S', max])
ret.append(['STD_S', hfh.c_round(graded_df['SCORE'].std(), 2)])
ret = hfh.pd.DataFrame(ret).T
ret.columns = ret.loc[0]
ret = ret.set_index(ret['DATE'])
return ret
else:
if f_path.find('_c.csv') == -1:
print(f_path +
" was not configured as _f with date information")
def create_upload_from_processed(c_file,
f_file,
path=None,
c_has_header=True,
to_csv=False):
#todo: decide if _c files have headers or not...
#todo: perhaps a different _x to indicate header or not... the only time I don't want a header is xCalibre...
if c_has_header:
c_df = hfh.get_df(c_file, header=0)
else:
c_df = hfh.get_df(c_file)
f_df = hfh.get_df(f_file, index_col=0)
stats_df = hfh.pd.DataFrame([])
stats_df['AccNum'] = c_df.iloc[:, 0]
graded_df = hr.grade_examination(f_df,
c_df,
grading_processed=True,
correct=1,
incorrect=0)
score = graded_df.sum(axis=1)
pbis = graded_df[graded_df.columns[0]].corr(score)
A = get_option_df(f_df, 'A')
B = get_option_df(f_df, 'B')
C = get_option_df(f_df, 'C')
D = get_option_df(f_df, 'D')
options = ['A', 'B', 'C', 'D']
dfs = [A, B, C, D]
counter = -1
N = ~f_df.isna()
N = N.sum()
N = N.reset_index(drop=True)
for option in options:
counter += 1
a_ret = []
b_ret = []
c_ret = []
df = dfs[counter]
for column in A.columns:
mask = df[column] == 1
mean_score = graded_df[mask].mean().mean()
c_ret.append(mean_score)
pbis = df[column].corr(score)
endorse_p = df[column].sum() / df.shape[0]
a_ret.append(pbis)
b_ret.append(endorse_p)
stats_df[option + '_r'] = hfh.pd.Series(a_ret, index=stats_df.index)
stats_df[option + '_p'] = hfh.pd.Series(b_ret, index=stats_df.index)
stats_df[option + '_m'] = hfh.pd.Series(c_ret, index=stats_df.index)
k_ret = []
for i in range(graded_df.shape[1]):
pbis = graded_df[graded_df.columns[i]].corr(score)
k_ret.append(pbis)
stats_df['K_r'] = hfh.pd.Series(k_ret, index=stats_df.index)
stats_df['KEY'] = c_df['Key']
stats_df['N'] = N
p = graded_df.mean(axis=0)
stats_df = stats_df.set_index('AccNum', drop=True)
stats_df['P'] = p
if path is None:
name = hfh.get_stem(f_file)[:-2] + '_P.csv'
else:
name = path + '/' + hfh.get_stem(f_file)[:-2] + '_P.csv'
stats_df = stats_df[[
'KEY', 'K_r', 'P', 'A_p', 'A_r', 'A_m', 'B_p', 'B_r', 'B_m', 'C_p',
'C_r', 'C_m', 'D_p', 'D_r', 'D_m', 'N'
]]
if to_csv:
stats_df.to_csv(name)
return stats_df