def median(headers, data):
'''Takes in a data object and list of headers. Returns the median of appropriate data.'''
cols = []
for h in headers:
cols.append(data.header2col[h])
types = data.get_types()
for c in cols:
if (types[c].strip() != 'numeric'):
print('Only numeric types allowed!')
return
D = data.columns_data(headers)
medians = []
for i in range(0,len(headers)):
element = np.median(D[:, i].astype(np.float))
medians.append(element)
return medians
def stdev(headers, data):
'''Take in a data object and list of headers. Returns the standard deviation of
appropriate data.'''
cols = []
for h in headers:
cols.append(data.header2col[h])
types = data.get_types()
for c in cols:
if (types[c].strip() != 'numeric'):
print('Only numeric types allowed!')
return
D = data.columns_data(headers) # this is a np matrix
deviations = []
for i in range(0,len(headers)):
element = np.std(D[:, i].astype(np.float))
deviations.append(element)
return deviations
def data_range(headers, data):
'''Takes in a Data object and a list of column headers and returns a list of 2-element
lists with the min and max values for each column. Only deals with numeric types.'''
cols = []
for h in headers:
cols.append(data.header2col[h])
types = data.get_types()
for c in cols:
if (types[c].strip() != 'numeric'):
print('Only numeric types allowed!')
return
D = data.columns_data(headers)
mins_maxes = []
for i in range(0,len(headers)):
element = [float(np.min(D[:, i].astype(np.float))),float(np.max(D[:, i].astype(np.float)))]
mins_maxes.append(element)
return mins_maxes
def normalize_columns_separately(headers, data):
'''Takes in a Data object and a list of headers. Returns a matrix with each columns normalized
so that the min is mapped to 0 and the max is mapped to 1.'''
cols = []
for h in headers:
cols.append(data.header2col[h])
types = data.get_types()
for c in cols:
if (types[c].strip() != 'numeric'):
print('Only numeric types allowed!')
return
D = data.columns_data(headers)
normcols_sep = D.copy()
DR = data_range(headers,data)
for i in range(0, len(DR)):
a = DR[i][0] #min
b = DR[i][1] #max
r = b - a
normcols_sep[:, i] = (D[:, i].astype(np.float) - a) / r
return normcols_sep
def normalize_columns_together(headers, data):
'''Takes in a Data object and a list of headers. Returns a matrix with columns normalized
so that the min across columns is mapped to 0 and the max across columns is mapped to 1.'''
cols = []
for h in headers:
cols.append(data.header2col[h])
types = data.get_types()
for c in cols:
if (types[c].strip() != 'numeric'):
print('Only numeric types allowed!')
return
D = data.columns_data(headers)
normcols_tog = D.copy()
DR = data_range(headers, data)
mins = []
maxes = []
for twople in DR:
mins.append(twople[0])
maxes.append(twople[1])
verymin = min(mins)
verymax = max(maxes)
R = verymax - verymin
normcols_tog[:, :] = (D[:, :].astype(np.float) - verymin) / R
return normcols_tog