*STATA TUTORIAL: MICROECONOMETRICS

*The objective of this do file is to show you how to use a do file for you analysis. 
*It will also contain some commands we will oversee during the lecture so that you can use them later.
*Please keep a copy of this do file and the slides in your H drives for future reference.

*A1
***********************
*Starting a do file

clear
	*clear deletes any previous datasets, variables, etc. that are in the current memory
set mem 50m
	*This tells the computer how much memory to allocate to stata. You will get an error messageif memory is insufficient
capture log close
	*This tells stata to close a log file if it is open (a file that records all output from your stat session)


*In general you will download your data using commands such as:
	* use "C:\database.dta"
	* insheet using "C:\database.csv"

*For the tutorial I will instead generate a dataset so that you can run some commands. 
	set obs 1000
	*This tells stata the database will have 1000 observations
	
*Slide: Dates in Stata are numbers
	*For the time series example:
	gen datevar=_n
	format datevar %td
	*browse
	format datevar %tm
	*browse
	tsset datevar

*Slide: What if date comes in two variables:
	clear
	set obs 1000
	gen month=_n
	gen year=_n+30

*Slide: Dickey Fuller 
	gen y=uniform()*100
	*Note: this is iid variable: should have no unit root
	tsset month
	dfuller y
	*We reject the unit root hypothesis
	drop y
	gen x=uniform()*100
	gen y=300*_n+uniform()*100
	dfuller y
	*We detected a unit root

*Slide: Cointegration
	gen z=0.8*y+uniform()*100
	dfuller y
	dfuller z
	reg z y
	predict res, r
	dfuller res
	

*Slide fixed effects: 
	clear
	set obs 1000

	gen x1=invnorm(uniform())*100
		*Our x1 variable is normal distributed with mean 0 and sd 100
	gen x2=invnorm(uniform())*100
		*Our x2 variable is normal distributed with mean 0 and sd 100
	gen alfa=int(_n/10)*100
		*This is the fixed effect. Every 10 observations share the same alfa
		*Note that alfa is uncorrelated with x1 or x2 this will be important below
	gen year= mod(_n,10)

	gen y=3*x1+5*x2+alfa+80*year+invnorm(uniform())*150
		*This is the true relationship between y and the x's plus some normal iid noise

*Traditional (cross section) regression in Stata:
	reg y x1 x2, r

***********************
	*Fixed effects regression using xtreg:
	xtset alfa year
	xtreg y x1 x2 ,fe r