def plotSlopeHists(data):
summed = util.sumBy(data,'CCG')
getPercents(summed)
slopes = findSlopes(summed)
pyplot.hist(slopes, bins = 50)
pyplot.title('Actual distribution of rate of change \n of diclophenac prescription',
fontsize = 18)
pyplot.xlabel('rate of change of percent diclofenac prescribed', fontsize = 14)
pyplot.ylabel('number of CCGs', fontsize = 14)
pyplot.axis([-.06, .08, 0, 38])
pyplot.show()
CCGs = data.to_dict(outtype='list')['CCG']
random.shuffle(CCGs)
data['CCG'] = CCGs
summed = util.sumBy(data,'CCG')
getPercents(summed)
slopes = findSlopes(summed)
pyplot.hist(slopes, bins = 50)
pyplot.title('Example distribution of rate of change \n of diclofenac prescription without \ngovernance or geographic effect',
fontsize = 18)
pyplot.xlabel('rate of change of percent diclofenac prescribed', fontsize = 14)
pyplot.ylabel('number of CCGs', fontsize = 14)
pyplot.axis([-.06, .08, 0, 38])
pyplot.show()
def plotFinalHists(data):
data['days_prescribed_diclofenac'] = data['days_prescribed_diclofenac'].map(lambda x: 0 if x!=x else x)
data['days_prescribed_naproxen'] = data['days_prescribed_naproxen'].map(lambda x: 0 if x!=x else x)
summed = util.sumBy(data,'CCG')
summed['perc'] = summed['days_prescribed_diclofenac']/(summed['days_prescribed_naproxen']+summed['days_prescribed_diclofenac'])
pyplot.hist(summed['perc'], bins = 100)
pyplot.title('Actual distribution of rate of diclophenac prescription',
fontsize = 18)
pyplot.xlabel('percent diclofenac prescribed', fontsize = 14)
pyplot.ylabel('number of CCGs', fontsize = 14)
pyplot.axis([0, 1, 0, 30])
pyplot.show()
CCGs = data.to_dict(outtype='list')['CCG']
random.shuffle(CCGs)
data['CCG'] = CCGs
summed = util.sumBy(data,'CCG')
summed['perc'] = summed['days_prescribed_diclofenac']/(summed['days_prescribed_naproxen']+summed['days_prescribed_diclofenac'])
pyplot.hist(summed['perc'], bins = 100)
pyplot.title('Example distribution of rate of diclofenac \nprescription without governance or geographic effect',
fontsize = 18)
pyplot.xlabel('percent diclofenac prescribed', fontsize = 14)
pyplot.ylabel('number of CCGs', fontsize = 14)
pyplot.axis([0, 1, 0, 30])
pyplot.show()
def run(self):
bfile = self.Config.append_dir("WriteRatioDatasetBrand")
gfile = self.Config.append_dir("WriteRatioDatasetGeneric")
ofile = self.Config.append_dir("WriteRatioDatasetOut")
for (brand, generic, outfile) in zip(bfile,gfile,ofile):
brandfile = self.loadDF(brand)
# if not brandfile:
# continue
genericfile = self.loadDF(generic)
# if not genericfile:
# continue
groupedbrand = util.sumBy(brandfile,[self.Config.keys['practice'],'postal code'])
groupedgeneric = util.sumBy(genericfile,[self.Config.keys['practice'],'postal code'])
output = pandas.DataFrame.merge(
groupedbrand,
groupedgeneric,
on=self.Config.keys['practice'],
how = 'outer',
suffixes =('brand','generic'))
items = self.Config.keys['items']
quan = self.Config.keys['quantity']
nic = self.Config.keys['nic']
parts = [[items,quan,nic],['brand','generic']]
drops = []
for a in parts[0]:
for b in parts[1]:
output[a+b] = output[a+b].map(lambda x: 0 if x!=x else x)
drops.append(a+b)
drops = drops+['INCLUDEbrand','GENERICbrand','INCLUDEgeneric',
'GENERICgeneric','postal codebrand','postal codegeneric']
output['postal code'] = output.apply(
lambda row: row['postal codegeneric']
if row['postal codebrand']!=row['postal codebrand']
else row['postal codebrand'],
axis=1)
output['sumitems'] = output[items+'brand']+output[items+'generic']
output['sumquantity'] = output[quan+'brand']+output[quan+'generic']
output['sumnic'] = output[nic+'brand']+output[nic+'generic']
output['ratioitems']= output[items+'brand']/output['sumitems']
output['ratioquantity'] = output[quan+'brand']/output['sumquantity']
output['rationic'] = output[nic+'brand']/output['sumnic']
output = output.drop(drops,axis=1)
output.to_csv(outfile, index = False)
def run(self):
infiles = self.Config.append_dir("OutCodeRatiosIn")
outfiles = self.Config.append_dir("OutCodeRatiosOut")
outcodes = self.loadDF('postcodes.csv')
if not outcodes:
return
for infile,outfile in zip(infiles,outfiles):
data = self.loadDF(infile)
if not data:
continue
labels = ['items','quantity','nic']
for a in labels:
data['tot'+a] = data['sum'+a]*data['ratio'+a]
print data
print data['postal code']
data['outcode'] = data['postal code'].map(lambda x: x.partition(' ')[0])
data = util.sumBy(data,'outcode')
for a in labels:
data['ratio'+a] = data['tot'+a]/data['sum'+a]
data.drop('tot'+a,axis=1)
data = pandas.DataFrame.merge(data,outcodes,on='outcode',how='left')
data.to_csv(outfile)
def calc_drug_over_time(Config):
folder = 'SepGeneric'
infiles = Config.append_dir(folder)
quantity = {}
nic = {}
for month in infiles:
df = loadDF(month)
month = month[-11:-4]
grouped = util.sumBy(df,Config.keys['bnf'])
for index,row in grouped.iterrows():
item =row[Config.keys['bnf']]
if item in quantity.keys():
quantity[item][month] = (row[Config.keys['quantity']])#,row[Config.keys['nic']])
nic[item][month] = (row[Config.keys['nic']])#,row[Config.keys['nic']])
else:
quantity[item] = {}
quantity[item][month] = (row[Config.keys['quantity']])#,row[Config.keys['nic']])
nic[item] = {}
nic[item][month] = (row[Config.keys['nic']])#,row[Config.keys['nic']])
return quantity,nic
def calc_drug_over_time(self, grouping = 'practice'):
folder = 'NSAIDSummed'
infiles = self.Config.append_dir(folder)
naproxen = {}
diclofenac = {}
for month in infiles:
df = self.loadDF(month)
if grouping == 'CCG':
df = util.sumBy(df,self.Config.keys['ccg'])
time.sleep(20)
month = month[-11:-4]
for index,row in df.iterrows():
if grouping == 'CCG':
item =row[self.Config.keys['ccg']]
else:
item = row[self.Config.keys['practice']]
if item in naproxen.keys():
naproxen[item][month] = row['days_prescribed_naproxen']
diclofenac[item][month] = row['days_prescribed_diclofenac']
else:
naproxen[item] = {}
naproxen[item][month] = (row['days_prescribed_naproxen'])
diclofenac[item] = {}
diclofenac[item][month] = (row['days_prescribed_diclofenac'])
return naproxen,diclofenac
def findSlopeMeanVar(base): summed = util.sumBy(base,'CCG') getPercents(summed) slopes = findSlopes(summed) return ts.MeanVar(slopes)
def run(self):
infiles = self.Config.append_dir("OutCodeDrugsIn")
outfiles = self.Config.append_dir("OutCodeDrugsOut")
outcodes = self.loadDF('postcodes.csv')
if not outcodes:
return
for infile,outfile in zip(infiles,outfiles):
data = self.loadDF(infile)
if not data:
continue
data['outcode'] = data['postal code'].map(lambda x: x.partition(' ')[0])
data = util.sumBy(data,['outcode','ChemID'])
items = self.Config.keys['items']
quan = self.Config.keys['quantity']
nic = self.Config.keys['nic']
cols = [items,quan,nic]
for col in cols:
data['sum'+col] = data[col+'_brand']+data[col+'_gen']
data['percent'+col]= data[col+'_brand']/data['sum'+col]
data = data.drop([col+'_brand',col+'_gen'],axis=1)
data = pandas.DataFrame.merge(data,outcodes,on='outcode',how='left')
data.to_csv(outfile, index=False)
def plotRegression(self):
lines = self.loadDF("Results/CostQuanCorrs_PPI.csv")
lines = lines.groupby("Group")
for name, group in lines:
group = group.groupby("Month")
for month, fits in group:
data = self.loadDF(name + "/" + month + ".csv")
data = util.sumBy(data, self.Config.keys["bnf"])
data["cost"] = data[Config.keys["nic"]] / data[Config.keys["quantity"]]
data = data.to_dict(outtype="list")
pp.scatter(data["cost"], data[self.Config.keys["items"]])
xs = [x / 20.0 for x in range(30)]
for index, row in fits.iterrows():
if row["transform"] == "linear":
ys = [x * row["slope"] + row["inter"] for x in xs]
pp.plot(xs, ys, "-.", label="linear")
if row["transform"] == "log":
ys = [math.exp(x * row["slope"] + row["inter"]) for x in xs]
pp.plot(xs, ys, "-", label="log")
pp.legend()
pp.title(name + " " + month)
pp.ylabel("items")
pp.xlabel("cost")
pp.yscale("log")
pp.show()
def plotAllTime(self):
lines = self.loadDF("Results/CostQuanCorrs_PPI_AllTime.csv")
lines = lines.groupby("Group")
for name, group in lines:
xs = []
ys = []
for month in self.Config.filenames:
data = self.loadDF(name + "/" + month)
data = util.sumBy(data, self.Config.keys["bnf"])
data["cost"] = data[Config.keys["nic"]] / data[Config.keys["quantity"]]
data = data.to_dict(outtype="list")
xs = xs + data["cost"]
ys = ys + data[self.Config.keys["items"]]
pp.scatter(xs, ys)
fxs = [x / 20.0 for x in range(30)]
for index, row in group.iterrows():
if row["transform"] == "linear":
fys = [x * row["slope"] + row["inter"] for x in fxs]
pp.plot(fxs, fys, "-.", label="linear")
if row["transform"] == "log":
fys = [math.exp(x * row["slope"] + row["inter"]) for x in fxs]
pp.plot(fxs, fys, "-", label="log")
pp.legend()
pp.title(name)
pp.ylabel("items")
pp.xlabel("cost")
pp.yscale("log")
pp.xscale("log")
pp.show()
def chemGenBrandComp(Config):
infiles = Config.append_dir("Summed_Chem_Gen", group="psych")
antipsych = pandas.read_csv("Criteria/antipsychotics.csv", index_col="chem code")
data = {}
for i, r in antipsych.iterrows():
data[r["name"]] = {"brand": [0 for i in range(len(infiles))], "gen": [0 for i in range(len(infiles))]}
mon = 0
for infile in infiles:
print "Loading", infile
df = pandas.read_csv(infile, index_col=False)
df = util.sumBy(df, ["chem code", "name", "generic"])
for i, r in df.iterrows():
if r["generic"] == 1:
gen = "gen"
else:
gen = "brand"
tot = r[Config.keys["items"]]
data[r["name"]][gen][mon] = tot
mon += 1
for key, value in data.items():
if np.sum(value["brand"]) > 0 or np.sum(value["gen"]) > 0:
pp.plot(value["brand"], "r*-", label="brand")
pp.plot(value["gen"], "bo-", label="generic")
pp.legend()
pp.title("Total prescriptions of " + key)
pp.ylabel("# prescriptions")
pp.xlabel("months since Jan 2012")
pp.show()
def run(self):
infiles = self.Config.append_dir("Sum_by_ccg_in", group = 'NSAID')
outfiles = self.Config.append_dir("Sum_by_ccg_out", group = 'NSAID')
for (infile,outfile) in zip(infiles,outfiles):
df = self.loadDF(infile)
data = util.sumBy(df,[self.Config.keys['ccg']])
data['sum'] = data['days_prescribed_naproxen']+data['days_prescribed_diclofenac']
data['percent']= data['days_prescribed_diclofenac']/data['sum']
data.to_csv(outfile, index = False)
def plotFinalCDF(data):
data['days_prescribed_diclofenac'] = data['days_prescribed_diclofenac'].map(lambda x: 0 if x!=x else x)
data['days_prescribed_naproxen'] = data['days_prescribed_naproxen'].map(lambda x: 0 if x!=x else x)
summed = util.sumBy(data,'CCG')
summed['perc'] = summed['days_prescribed_diclofenac']/(summed['days_prescribed_naproxen']+summed['days_prescribed_diclofenac'])
cdf1 = ts.MakeCdfFromList(summed['perc'])
pmf1 = ts.EstimatedPdf(summed['perc']).MakePmf([x/50.0 for x in range(51)])
CCGs = data.to_dict(outtype='list')['CCG']
random.shuffle(CCGs)
data['CCG'] = CCGs
summed = util.sumBy(data,'CCG')
summed['perc'] = summed['days_prescribed_diclofenac']/(summed['days_prescribed_naproxen']+summed['days_prescribed_diclofenac'])
cdf2 = ts.MakeCdfFromList(summed['perc'])
pmf2 = ts.EstimatedPdf(summed['perc']).MakePmf([x/50.0 for x in range(51)])
pyplot.subplot(1,2,1)
xs,ys = cdf1.Render()
pyplot.plot(xs,ys,'r-',linewidth=2,label = 'Actual')
xs,ys = cdf2.Render()
pyplot.plot(xs,ys,'b--',linewidth=2,label = 'No CCG correlation')
pyplot.title('CDF')
# pyplot.title('Actual distribution of percent diclofenac prescribed\ncompared to distribution with no CCG correlation',
# fontsize = 18)
pyplot.xlabel('percent diclofenac prescribed', fontsize = 14)
pyplot.ylabel('fraction of CCGs at or below percentage', fontsize = 14)
pyplot.legend(loc=4)
pyplot.axis([0, 1, 0, 1])
pyplot.subplot(1,2,2)
xs,ys = pmf1.Render()
pyplot.plot(xs,ys,'r-',linewidth=2,label = 'Actual')
xs,ys = pmf2.Render()
pyplot.plot(xs,ys,'b--',linewidth=2,label = 'No CCG correlation')
pyplot.title('PDF')
# pyplot.title('Actual distribution of percent diclofenac prescribed\ncompared to distribution with no CCG correlation',
# fontsize = 18)
pyplot.xlabel('percent diclofenac prescribed', fontsize = 14)
pyplot.ylabel('probability density', fontsize = 14)
pyplot.legend(loc=1)
pyplot.show()
def run(self):
na_infiles = self.Config.append_dir("Sum_by_practice_in_nap", group='NSAID')
dic_infiles = self.Config.append_dir("Sum_by_practice_in_dic", group = 'NSAID')
outfiles = self.Config.append_dir("Sum_by_practice_out", group = 'NSAID')
for (na_infile, dic_infile, outfile) in zip(na_infiles,dic_infiles, outfiles):
na_df = self.loadDF(na_infile)
dic_df = self.loadDF(dic_infile)
grouped_na = util.sumBy(na_df,[self.Config.keys['practice'], self.Config.keys['ccg'],self.Config.keys['pct']])
grouped_dic = util.sumBy(dic_df,[self.Config.keys['practice'], self.Config.keys['ccg'],self.Config.keys['pct']])
output = pandas.DataFrame.merge(
grouped_na,
grouped_dic,
on=[self.Config.keys['practice'], self.Config.keys['ccg'], self.Config.keys['pct']],
how = 'outer',
suffixes =('_naproxen','_diclofenac'))
print output
output.to_csv(outfile, index = False)
def plotSlopeCDF(data):
summed = util.sumBy(data,'CCG')
getPercents(summed)
slopes = findSlopes(summed)
cdf1 = ts.MakeCdfFromList(slopes)
pmf1 = ts.EstimatedPdf(slopes).MakePmf([x/200.0 for x in range(-12,21)])
CCGs = data.to_dict(outtype='list')['CCG']
random.shuffle(CCGs)
data['CCG'] = CCGs
summed = util.sumBy(data,'CCG')
getPercents(summed)
slopes = findSlopes(summed)
cdf2 = ts.MakeCdfFromList(slopes)
pmf2 = ts.EstimatedPdf(slopes).MakePmf([x/200.0 for x in range(-12,21)])
pyplot.subplot(1,2,1)
xs,ys = cdf1.Render()
pyplot.plot(xs,ys,'r-',linewidth=2,label = 'Actual')
xs,ys = cdf2.Render()
pyplot.plot(xs,ys,'b--',linewidth=2,label='No CCG correlation')
pyplot.xlabel('rate of change of percent diclofenac prescribed', fontsize = 14)
pyplot.ylabel('fraction of CCGs at or below rate', fontsize = 14)
pyplot.legend(loc=4)
pyplot.subplot(1,2,2)
xs,ys = pmf1.Render()
pyplot.plot(xs,ys,'r-',linewidth=2,label = 'Actual')
xs,ys = pmf2.Render()
pyplot.plot(xs,ys,'b--',linewidth=2,label = 'No CCG correlation')
# pyplot.title('Actual distribution of rate of change\nof diclophenac prescription compared to\ndistribtuion without any CCG correlation ',
# fontsize = 18)
pyplot.xlabel('rate of change of percent diclofenac prescribed', fontsize = 14)
pyplot.ylabel('probability density', fontsize = 14)
pyplot.legend(loc=1)
pyplot.show()
def plotCostVQuantity(filenames, Config):
for datafile in datafiles:
print datafile
df = pandas.read_csv(datafile)
df = util.sumBy(df, Config.keys["bnf"])
df["ITEM COST"] = df[Config.keys["nic"]] / df[Config.keys["quantity"]]
plotTransform(df, "ITEM COST", Config.keys["quantity"], np.log, "Log")
inter, slope = findCorr(df, "ITEM COST", Config.keys["quantity"], np.log)
ys = [inter + slope * x for x in np.log(df["ITEM COST"])]
pp.scatter(np.log(df["ITEM COST"]), ys, color="red")
pp.show()
def runAllTime(self, items_cutoff=0, cost_cutoff=0):
out = {}
infolders = self.Config.directories["CorrsIn"]
for folder in infolders:
infiles = self.Config.append_dir(folder)
xs = []
ys = []
for infile in infiles:
data = self.loadDF(infile)
data = util.sumBy(data, self.Config.keys["bnf"])
data["cost"] = data[self.Config.keys["nic"]] / data[self.Config.keys["quantity"]]
data = data[data[self.Config.keys["items"]] > items_cutoff]
data = data[data["cost"] > cost_cutoff]
data = data.to_dict(outtype="list")
xs = xs + data["cost"]
ys = ys + data[self.Config.keys["items"]]
corr = ts.SpearmanCorr(xs, ys)
pVal = self.PValue(xs, ys, actual=corr, n=1000)
sdev, serr, inter, slope = self.regress(xs, ys)
out["Group"] = out.get("Group", []) + [folder]
out["transform"] = out.get("transform", []) + ["linear"]
out["Scorr"] = out.get("Scorr", []) + [corr]
out["p"] = out.get("p", []) + [pVal]
out["Stan Dev"] = out.get("Stan Dev", []) + [sdev]
out["Stan Err"] = out.get("Stan Err", []) + [serr]
out["inter"] = out.get("inter", []) + [inter]
out["slope"] = out.get("slope", []) + [slope]
sdev, serr, inter, slope = self.regress(xs, ys, ylog=True)
out["Group"] = out.get("Group", []) + [folder]
out["transform"] = out.get("transform", []) + ["log"]
out["Scorr"] = out.get("Scorr", []) + [corr]
out["p"] = out.get("p", []) + [pVal]
out["Stan Dev"] = out.get("Stan Dev", []) + [sdev]
out["Stan Err"] = out.get("Stan Err", []) + [serr]
out["inter"] = out.get("inter", []) + [inter]
out["slope"] = out.get("slope", []) + [slope]
# pp.scatter(xs,ys)
# pp.yscale('log')
# pp.show()
data = pandas.DataFrame(out, index=None)
data.to_csv("Results/CostQuanCorrs_PPI_AllTime.csv")
def run(self):
infiles = self.Config.append_dir("MakeDrugPairsIn")
outfiles = self.Config.append_dir("MakeDrugPairsOut")
for (infile,outfile) in zip(infiles,outfiles):
data = self.loadDF(infile)
if not data:
continue
data['ChemID'] = data[self.Config.keys['bnf']].map(lambda x: x[0:9])
data = util.sumBy(data,[self.Config.keys['practice'],'ChemID',self.Config.keys['gen'],'postal code'])
grouped = pandas.groupby(data,self.Config.keys['gen'])
data = pandas.merge(grouped.get_group(1.0),grouped.get_group(0.0),
on =[self.Config.keys['practice'],'ChemID'],
left_index=False,
right_index = False,
how = 'outer',
sort = False,
suffixes = ('_gen','_brand'))
for col in data.columns.values.tolist():
data[col] = data[col].map(lambda x: 0 if x!=x else x)
data['postal code'] = data.apply(
lambda row: row['postal code_gen']
if row['postal code_brand']!=row['postal code_brand']
else row['postal code_brand'],
axis=1)
items = self.Config.keys['items']
quan = self.Config.keys['quantity']
nic = self.Config.keys['nic']
cols = [items,quan,nic]
for col in cols:
data['sum'+col] = data[col+'_brand']+data[col+'_gen']
data['percent'+col]= data[col+'_brand']/data['sum'+col]
data = data.drop(['INCLUDE_gen','INCLUDE_brand',
'GENERIC_gen','GENERIC_brand',
'postal code_gen','postal code_brand'],
axis=1)
data.to_csv(outfile, index = False)
def stackPlot(Config):
infiles = Config.append_dir("Ingest_out", group="psych")
antipsych = pandas.read_csv("Criteria/antipsychotics.csv", index_col="chem code")
data = {}
for i, r in antipsych.iterrows():
data[r["name"]] = []
for infile in infiles:
print "Loading", infile
df = pandas.read_csv(infile)
df = util.sumBy(df, ["chem code", "name"])
df = df.set_index("chem code")
for i, r in antipsych.iterrows():
try:
# relies on chem code being index
tot = df.loc[i][Config.keys["items"]]
except KeyError:
tot = 0
if tot != tot:
tot = 0
data[r["name"]].append(tot)
prev = None
lines = []
legends = []
for key, value in sorted(data.items(), key=lambda x: numpy.mean(x[1])):
if prev:
prev = sumLists(prev, value)
else:
prev = value
if numpy.mean(value) < 1000:
pp.plot(prev)
else:
lines = pp.plot(prev) + lines
legends = [key] + legends
pp.legend(lines, legends)
pp.title("Cumulative presecriptions of antipsychotics")
pp.ylabel("# prescriptions")
pp.xlabel("months since Jan 2012")
pp.show()
def run(self):
for files, dirname in zip(self.inputs,self.inDirs):
dists = {}
for datafile in files:
date = datafile[-11:-4]
df = self.loadDF(datafile)
df = util.sumBy(df,self.Config.keys['bnf'])
for distFun in self.distFuns:
res = distFun[1](df)
res['Date'] = date
dists[distFun[0]] = dists.get(distFun[0],[]) + [res]
for distFun in self.distFuns:
outfile = self.outDir+'/'+dirname+'_'+distFun[0]+'_SummaryStats.csv'
df = pandas.DataFrame(dists[distFun[0]])
df.to_csv(outfile,index=None)
def run(self):
infiles = self.Config.append_dir("Sum_by_practice_out", group='NSAID')
outfiles = self.Config.append_dir("NSAIDGov")
pctNames = pandas.read_csv("Criteria/pctCodeToName.csv")
for infile, outfile in zip(infiles,outfiles):
df = self.loadDF(infile)
output = util.sumBy(df,[self.Config.keys['ccg'],self.Config.keys['pct']])
dd = output['days_prescribed_diclofenac']
dn = output['days_prescribed_naproxen']
output['percent'] = dd/(dd+dn)
out = pandas.DataFrame.merge(
output,
pctNames,
left_on=self.Config.keys['pct'],
right_on = 'code',
how = 'outer')
out.to_csv(outfile, index = False)
def run(self):
infiles = self.Config.append_dir("AllDrugsIn")
outfiles = self.Config.append_dir("AllDrugsOut")
for infile,outfile in zip(infiles,outfiles):
data = self.loadDF(infile)
if not data:
continue
data = util.sumBy(data,['ChemID'])
items = self.Config.keys['items']
quan = self.Config.keys['quantity']
nic = self.Config.keys['nic']
cols = [items,quan,nic]
for col in cols:
data['sum'+col] = data[col+'_brand']+data[col+'_gen']
data['percent'+col]= data[col+'_brand']/data['sum'+col]
data = data.drop([col+'_brand',col+'_gen'],axis=1)
data.to_csv(outfile, index=False)
# data[row['CCG']] = [row['percent']] + data.get(row['CCG'],[])
# for key, value in data.items():
# pyplot.plot(value)
# '''
# pyplot.plot(data['mean'], label = 'average')
# pyplot.plot(data['dev_up'],'.', label = '1 std dev up')
# pyplot.plot(data['dev_down'],'.', label = '1 std dev down')
# pyplot.legend()
# pyplot.title('Averge percent diclofenac over all CCGs (Jan 2012 to Oct 2013)')
# pyplot.ylabel('Percent diclofenac')
# pyplot.xlabel('Months since Jan 2012')
# pyplot.show()
infile = Config.nsaid_directories['Ingest_out_nap']+'/Oct2013.csv'
data = pandas.read_csv(infile)
data = util.sumBy(data,Config.keys['bnf'])
plotLogCDF(data,
'days_prescribed',
'Log(x) CDF of days prescribed for NSAID preparations (Oct 2013)',
'','Days prescribed')
# infile = Config.directories['Ingest_out']+'/Oct2013.csv'
# data = pandas.read_csv(infile)
# data = util.sumBy(data,Config.keys['bnf'])
# plotLogCDF(data,
# Config.keys['items'],
# 'Log(x) CDF of items for all drug preparations (Oct 2013)',
# '','Prescriptions filled')
# data['cost'] = data[Config.keys['nic']]/data[Config.keys['quantity']]
# plotLogCDF(data,
def process(self, df):
return util.sumBy(df, ["chem code", "name", "CCG", "PCT", "generic"])
mkPlotMedianMean(data2,"blue",name=data2label)
def mkPlotVersus(df,x,y,**args):
data = df.to_dict(outtype='list')
pp.scatter(data[x],data[y],**args)
if __name__ =="__main__":
Config = config.Config()
'''PPI items vs. cost'''
for brand, gen in zip(Config.append_dir('SepBrand'),Config.append_dir('SepGeneric')):
df_Brand = pandas.read_csv(brand)
df_Gen = pandas.read_csv(gen)
df_Brand = util.sumBy(df_Brand,Config.keys['bnf'])
df_Gen = util.sumBy(df_Gen,Config.keys['bnf'])
df_Brand['cost'] = df_Brand[Config.keys['nic']]/df_Brand[Config.keys['quantity']]
df_Gen['cost'] = df_Gen[Config.keys['nic']]/df_Gen[Config.keys['quantity']]
mkPlotVersus(df_Brand,'cost',Config.keys['items'],marker='x',s=35,color="red",label="Brand")
mkPlotVersus(df_Gen,'cost',Config.keys['items'],marker = '*',s=35,color="blue",label="Generic")
pp.title(brand)
# pp.title('PPIs (Oct 2013)')
pp.ylabel('Number of prescriptions for drug')
pp.xlabel('Drug Cost')
# pp.legend()
# pp.yscale('log')
pp.show()
def makeDrugSums(filename,Config): '''returns dataframe from file, aggregated by bnf code''' df = pandas.read_csv(filename) df = util.sumBy(df,Config.keys['bnf']) return df
return data
if __name__ == "__main__":
Config = config.TestConfig()
'''
data = makeDrugSums('CompressedData/Oct2013.csv',Config)
data['cost'] = data[Config.keys['nic']]/data[Config.keys['quantity']]
'''
data = pandas.read_csv('RatioDataset/Oct2013.csv')
labels = ['items','quantity','nic']
for a in labels:
data['tot'+a] = data['sum'+a]*data['ratio'+a]
#data['outcode'] = data['postal code'].map(lambda x: x.partition(' ')[0])
data = util.sumBy(data,Config.keys['practice'])
for a in labels:
data['ratio'+a] = data['tot'+a]/data['sum'+a]
data.drop('tot'+a,axis=1)
plotLogNormal(data,'ratioitems',res=4000)
plotLogNormal(data,'rationic',res = 4000)
'''
print 'All Drugs (Cost)'
df = makeDrugSums("CompressedData/Oct2013.csv", Config)
df['cost'] = df[Config.keys['nic']]/df[Config.keys['quantity']]
util.plotLogNormal(df,'cost')
def findFinalMeanVar(data): summed = util.sumBy(data,'CCG') summed['perc'] = summed['days_prescribed_diclofenac']/(summed['days_prescribed_naproxen']+summed['days_prescribed_diclofenac']) return ts.MeanVar(summed['perc'])
