st: RE: CDF log-normal and histogram

[Nick Cox <n.j.cox@durham.ac.uk>]

Not the graph I would use in that territory. 

See -lognfit- (Stephen Jenkins, SSC) and -qlogn- from -qpfit- (SSC) for more critical choices. 

But in terms of what you ask, I don't know what "doesn't work" means in your case.

Consider

sysuse auto, clear
gen ln_price = ln(price) 
su ln_price 

distplot price, ///
addplot(function normal((ln(x) - `r(mean)')/`r(sd)'), lp(dash) ra(1 16000)) ///
legend(order(1 "delicious data" 2 "mad model")) yla(, ang(h))

If you want to fit the distribution in other ways, plug in your parameter estimates accordingly. 

Here's a stab at a histogram too: 

sysuse auto, clear
gen ln_price = ln(price) 
su ln_price 

distplot price, ///
addplot(function normal((ln(x) - `r(mean)')/`r(sd)'), lp(dash) ra(1 16000) ///
|| histogram price, fraction bfcolor(none) width(2000)) ///
legend(order(1 "delicious data" 2 "mad model")) yla(, ang(h))

In terms of giving references to user-written stuff as requested:

-distplot- is from SJ. 
-dpplot- is from SJ. 

Authors not named are me. 

Nick 
n.j.cox@durham.ac.uk 

Beatrice Crozza

I want to have a graph which combine the histogram of frequencies for
my variable and on the same graph an empirical CDF of data and a
fitted CDF (log-normal)

I can do the histogram in this way:
histogram var1, freq

To see if my variable follows a log-normal I do:
dpplot var1, dist(log-normal)

The CDF of my variable in this way:
distplot var1,

How can I combine the histogram with the empirical CDF and the fitted
CDF (log-normal)?

I tried with twoway but it doesn't work.

Thank you very much for your help.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/