def idbar2(acertos,acertos2,maxitems=10,fig=None,ax=None):
''
itemstats, teststats = stats(acertos)
itemstats2, teststats2 = stats(acertos2)
id27 = itemstats['id25']
id279 = itemstats2['id25']
iddf = pandas.DataFrame(id27,index=range(1,len(id27)+1))
iddf = iddf.sort(columns=0)
iddf = iddf[0:maxitems]
iddf9 = pandas.DataFrame(id279,index=range(1,len(id279)+1))
iddf9 = iddf9.sort(columns=0)
iddf9 = iddf9[0:maxitems]
if not fig:
fig = plt.figure()
if not ax:
ax = fig.add_subplot(111)
ax.set_title(u"Índice de Discriminação (quartis)")
fig,ax = orderedfig(iddf.index,iddf[0],iddf9.index,iddf9[0],maxitems,fig,ax)
ax.set_xlabel(u"")
ax.set_ylim(-0.01,0.15)
ax.text(0.5,-0.1,u"Item",clip_on=False,transform = ax.transAxes,ha='center')
return fig,ax
def itemfbar2(acertos,acertos2,maxitems=10,fig=None,ax=None):
''
itemstats, teststats = stats(acertos)
itemf = itemstats['itemf']
itemstats2, teststats2 = stats(acertos2)
itemf2 = itemstats2['itemf']
itemfdf = pandas.DataFrame(itemstats['itemf'],index=range(1,len(itemf)+1))
itemfdf = itemfdf.sort(columns=0)
itemfdf = itemfdf[0:maxitems]
itemfdf2 = pandas.DataFrame(itemstats2['itemf'],index=range(1,len(itemf2)+1))
itemfdf2 = itemfdf2.sort(columns=0)
itemfdf2 = itemfdf2[0:maxitems]
if not fig:
fig = plt.figure()
if not ax:
ax = fig.add_subplot(111)
ax.set_title(u"Índice de Dificuldade")
ax.set_ylabel(u"Fração dos alunos que acertaram a questão")
ax.set_ylim(0,0.25)
fig,ax = orderedfig(itemfdf.index,itemfdf[0],itemfdf2.index,itemfdf2[0],maxitems,fig,ax)
ax.text(0.5,-0.1,u"Itens em ordem de dificuldade",clip_on=False,transform = ax.transAxes,ha='center')
return fig,ax
def iccfitgraph(df,acertos,fig=None):
''
if not fig:
fig = plt.figure()
#fig.suptitle(u"Parámetros dos fits logisticos")
ax1 = fig.add_subplot(211)
ax1.set_title(u"Dificuldade")
ax2 = fig.add_subplot(212)
ax2.set_title(u"Discriminição")
itemstats, teststats = stats(acertos,'nota')
iccfitsparam = itemstats['iccfitsparam']
itemds = [p[4] for p in iccfitsparam]
err = [p[5] for p in iccfitsparam]
x = np.arange(1,len(itemds)+1)
ax1.errorbar(x,itemds,yerr=err,fmt='o')
ax1.set_ylim(0,1400)
ax1.set_xlim(0,48)
ax1.set_xticklabels([])
itemn = [p[2] for p in iccfitsparam]
err = [p[3] for p in iccfitsparam]
x = np.arange(1,len(itemn)+1)
ax2.errorbar(x,itemn,yerr=err,fmt='o')
ax2.set_xlabel(u"Questão")
#ax1.set_ylim(0,1000)
ax2.set_xlim(0,48)
ax2.set_xticks([1,5,10,15,20,25,30,35,40,45])
return fig
def tpmfitgraph(df,acertos,qn,fig=None,ax=None):
''
if not fig:
fig = plt.figure()
if not ax:
ax = fig.add_subplot(111)
ax.set_title(u"Curva Característica do Item "+str(qn)+" (+ 3PL fit)")
itemstats, teststats = stats(acertos,hs = 'notapadrao',df=df)
hscale = itemstats['hscale']
icc = itemstats['icc'][qn-1]
hbin = icc[:,0]
nbin = icc[:,1]
acertos_no_bin = icc[:,2]
prob = icc[:,3]
err = icc[:,4]
ax.errorbar(hbin,prob,yerr=err,fmt='o')
x = np.linspace(0.9*min(hscale),1.1*max(hscale),200)
# p = invlogit(const+nota*x)
# ax.plot(x,p,'g-')
ax.set_ylim(0,1)
ax.set_yticks([0,0.5,1])
ax.set_xlabel(u"Escore Enem Padronizada")
ax.set_ylabel(u"Probabilidade")
return fig, ax
def iccgraph(df,acertos,qn,hs = "scores",fig=None,ax=None):
''
if not fig:
fig = plt.figure()
if not ax:
ax = fig.add_subplot(111)
ax.set_title(u"Curva Característica do Item "+str(qn))
itemstats, teststats = stats(acertos,hs = hs,df=df)
hscale = itemstats['hscale']
icc = itemstats['icc'][qn-1]
hbin = icc[:,0]
nbin = icc[:,1]
acertos_no_bin = icc[:,2]
prob = icc[:,3]
err = icc[:,4]
const,sconst,nota,snota,itemd,sitemd = itemstats['iccfitsparam'][qn-1]
ax.errorbar(hbin,prob,yerr=err,fmt='o')
x = np.linspace(0.9*min(hscale),1.1*max(hscale),200)
p = invlogit(const+nota*x)
ax.plot(x,p,'g-')
ax.set_ylim(0,1)
ax.set_yticks([0,0.5,1])
if hs == 'scores':
ax.set_xlabel(u"Acertos")
else:
ax.set_xlabel(u"Escore Enem")
ax.set_ylabel(u"Probabilidade")
return fig, ax
def resvec(df,rescol,gabcol,hscale=None):
''
dmap = {'A':1,
'B':2,
'C':3,
'D':4,
'E':5,
'.':'NA',
'*':'NA'
}
tonumbers = lambda x: dmap[x]
res = df[rescol]
gab = df[gabcol]
l = []
l1 = []
for rvec,gvec in zip(res,gab):
l.append([1 if x==y else 0 for x,y in zip(rvec,gvec)])
l1.append(map(tonumbers,rvec))
a = np.array(l)
an = np.array(l1)
df['res'] = list(a)
df['gab'] = gab
df['ressum'] = a.sum(axis=1)
df['resstd'] = a.std(axis=1)
itemstats, teststats = stats(a,hscale,df)
return df, itemstats, teststats, a, an
def ltmgrid(df,acertos,ncols=5,nrows=9,fig=None):
''
if not fig:
fig = plt.figure()
qn = 1
provid = df['ID_PROVA_CN'].values[0]
for row in range(nrows):
for col in range(ncols):
ax = plt.subplot2grid((nrows,ncols),(row,col))
itemstats, teststats = stats(acertos,hs = 'notapadrao',df=df)
hscale = itemstats['hscale']
icc = itemstats['icc'][qn-1]
hbin = icc[:,0]
nbin = icc[:,1]
acertos_no_bin = icc[:,2]
prob = icc[:,3]
err = icc[:,4]
ax.errorbar(hbin,prob,yerr=err,fmt='o')
x = np.linspace(0.9*min(hscale),1.1*max(hscale),200)
ltmdif,ltmdisc = ltmfitparams(provid)
a = ltmdisc['value'][qn]
b = ltmdif['value'][qn]
p = invlogit(1.0*a*(x-b))
ax.plot(x,p,'k-')
ax.set_ylim(0,1)
ax.set_yticks([0,0.5,1])
qn += 1
fig.subplots_adjust(left=0.1,right=0.95,bottom=0.05,top=0.9,wspace=0.4,hspace=0.4)
return fig
def itemfbar(acertos,acertos2,order=True,fig=None,ax=None):
''
itemstats, teststats = stats(acertos)
itemf = itemstats['itemf']
itemstats2, teststats2 = stats(acertos2)
itemf2 = itemstats2['itemf']
if order:
itemfdf = pandas.DataFrame(itemstats['itemf'],index=range(1,len(itemf)+1))
itemfdf = itemfdf.sort(columns=0)
itemfdf2 = pandas.DataFrame(itemstats2['itemf'],index=range(1,len(itemf2)+1))
itemfdf2 = itemfdf2.sort(columns=0)
if not fig:
fig = plt.figure()
if not ax:
ax = fig.add_subplot(111)
ax.set_title(u"Número do item em ordem de dificuldade")
width = 0.5
if order:
ax.bar(np.arange(1,len(itemf)+1),itemfdf[0],width=width,align='center',color='b',label="ENEM 2010")
ax.bar(np.arange(1+width,len(itemf)+1+width),itemfdf2[0],width=width,align='center',color="g",label="ENEM 2009")
else:
ax.bar(np.arange(1,len(itemf)+1),itemf,width=width,align='center')
# ax.plot([0,len(itemf)+1],[0.2,0.2],'b',label="Chute",linewidth=2)
# ax.plot([0,len(itemf)+1],[itemf.mean(),itemf.mean()],'k',label=u"Média",linewidth=2)
m10 = itemf.mean()
m9 = itemf2.mean()
ax.annotate(u"Médias",xy=(0,m10),xytext=(10,m10),color='k',arrowprops=dict(facecolor='b',width=2,shrink=0.05))
ax.annotate("",xy=(0,m9),xytext=(10,m9),color='g',arrowprops=dict(facecolor='g',width=2,shrink=0.05))
ax.annotate("Chute",xy=(0,0.2),xytext=(10,0.25),color='k',arrowprops=dict(facecolor='grey',width=1,shrink=0.05))
ax.legend(loc=2)
ax.set_ylabel(u"Fração dos alunos que acertaram a questão")
ax.set_xlabel(u"Item")
ax.set_ylim(0,1)
ax.set_xlim(0,len(itemf)+1)
if order:
ax.set_xticks([])
else:
ax.set_xticks([1,5,10,15,20,25,30,35,40,45])
return fig,ax
def idbar(acertos,acertos2,order=True,fig=None,ax=None):
''
itemstats, teststats = stats(acertos)
itemstats2, teststats2 = stats(acertos2)
#id50 = itemstats['id50']
id27 = itemstats['id27']
id279 = itemstats2['id27']
if order:
iddf = pandas.DataFrame(id27,index=range(1,len(id27)+1))
iddf = iddf.sort(columns=0)
iddf9 = pandas.DataFrame(id279,index=range(1,len(id279)+1))
iddf9 = iddf9.sort(columns=0)
if not fig:
fig = plt.figure()
if not ax:
ax = fig.add_subplot(111)
ax.set_title(u"Índice de Discriminação")
width = 0.5
if order:
ax.bar(np.arange(1,len(id27)+1),iddf[0],width=width,align='center',color='b',label=u'2010')
ax.bar(np.arange(1+width,len(id279)+1+width),iddf9[0],width=width,align='center',color='g',label="2009")
else:
ax.bar(np.arange(1,len(id27)+1),id27,width=width,align='center',color='b',label=u'diferença de acertos entre os piores e melhores 27%')
#ax.text(u'diferença de acertos entre os piores e melhores 27%')
ax.legend(loc="upper left")
ax.set_xlabel(u"Itens em ordem de discriminação")
ax.set_ylim(-0.05,0.4)
ax.set_xlim(0,len(id27)+1)
if order:
ax.set_xticks([])
else:
ax.set_xticks([1,5,10,15,20,25,30,35,40,45])
return fig,ax
def ltmfitgraph(df,acertos,qn1,qn2,fig=None):
'Só usar com prova 89 ou 49!'
provid = df['ID_PROVA_CN'].values[0]
if not fig:
fig = plt.figure()
fig.suptitle(u"Curvas Características dos Itens "+str(qn1)+" e "+str(qn2)+" (+ ajuste 2PL TRI)")
ax1 = fig.add_subplot(121)
ax2 = fig.add_subplot(122)
for qn,ax in [(qn1,ax1),(qn2,ax2)]:
itemstats, teststats = stats(acertos,hs = 'notapadrao',df=df)
hscale = itemstats['hscale']
icc = itemstats['icc'][qn-1]
hbin = icc[:,0]
nbin = icc[:,1]
acertos_no_bin = icc[:,2]
prob = icc[:,3]
err = icc[:,4]
ax.errorbar(hbin,prob,yerr=err,fmt='o')
x = np.linspace(0.9*min(hscale),1.1*max(hscale),200)
ltmdif,ltmdisc = ltmfitparams(provid)
a = ltmdisc['value'][qn]
b = ltmdif['value'][qn]
p = invlogit(1.0*a*(x-b))
ax.plot(x,p,'k-')
ax.set_ylim(0,1)
ax.set_yticks([0,0.5,1])
ax.set_xlabel(u"Escore Enem Padronizada")
ax.set_ylabel(u"Probabilidade")
ax2.set_ylabel('')
ax1.text(0.05,0.9,'Q'+str(qn1),transform = ax1.transAxes,fontsize='medium',weight='bold')
ax2.text(0.05,0.9,'Q'+str(qn2),transform = ax2.transAxes,fontsize='medium',weight='bold')
return fig