webuse hanley
 
tabulate disease
 
by disease, sort : tabulate rating

graph bar (count), over(rating) by(disease)



* cut point 4 for positivity
*   T+ = rating>=4
*   T- = rating< 4

gen test4="NEGATIVE"
replace test4="POSITIVE" if rating>=4
tab rating test4
sort rating
order test4, after(rating)
tabulate test4 disease, column

gen test3="NEGATIVE"
replace test3="POSITIVE" if rating>=3
tab rating test3
sort rating
order test3, before(test4)
tabulate test3 disease, column



***** 1) Use the roctab command to obtain descriptive statistics for ROC analysis
roctab disease rating
 roctab disease rating, detail graph


 
***** 2) Define a new rating newrating = rating + 100
gen newrating=rating+100

***** Compare the ROC analysis of rating and newrating 
roctab disease rating
roctab disease newrating

*** ROC curves are equal because with a constant shift in the
*** biomarker values (+100) we obtain the same relative frequencies in
*** diseased and non diseased. As a consequence sensitivity and 
*** specificity will be equal 


 help  cutpt
cutpt disease rating, noadjust
roctab disease rating, graph msymbol(none) addplot(scatteri `e(sens)' `=1 - e(spec)') legend(label(3 "Cutpoint"))





**************** confidence interval for prevalence *****************
ci proportions disease, wald

 display 51/109
display .46788991+1.96*sqrt(.46788991*(1-0.46788991)/109)
display .46788991-1.96*sqrt(.46788991*(1-0.46788991)/109)

**************** confidence interval for sensitivity*****************
tabulate test4 disease, column
gen test4_bin=0
replace test4_bin=1 if rating>=4


ci proportions test4_bin if disease==1, wald