/*--------------------------------------------
HOW TO USE THIS FILE

EVERYTHING INSIDE (STAR SLASH)  AND (SLASH STAR)
 IS TAKEN AS A COMMENT 
open it in the Stata do file editor

Run one line or group of lines at a time by
either highlighting them and going to DO in the tools window
(Run from this window will do it but give you no output)
Or press Ctrl^D 

/*--------------------------------------------
Analyse  SHS data in Stata Exemplar 2
First read in the file

set the survey description
and run the svydes to get a description of it

use "C:\Documents and Settings\gillian raab\My Documents\aprojects\peas\web\exemp2\data\ex2.dta", clear

--------------------------------------------------*/
svyset [pwei=ind_wt],psu(psu) strata(stratum)
svydes

/*-------------------------------------------------
First simple proportions of INTERNET USE

We give the code for this below, but in Stata you can 
Also get a dialogue box by searching for the commands in the help menu
When you complete this the code will be generated for you
------------------------------------------------*/
svyprop intuse
/*------------------------------------------------------
 svyprop does not give design effects so must treat proportion as
  the mean of a 0/1 variable to get the design effects

and we can also get them for subgroups
 -------------------------------------------------------------*/
svymean intuse , ci deff
svymean intuse, deff deft  by(sex)
/*------------------------------------------------------------
Now getting proportions of internet hours use for users
--------------------------------------------------------------*/
svymean rc5
/*------------------------------------------------------------
this fails because of lonely PSUs due to missing values
------------------------------------------------------------*/
svydes if intuse==1
/*---------------------------------------------------
This identifies them but there is 
no easy way round it except by setting it up as a 
regression

RESULTS FOR TABLE 2.4  IN EXEMPLAR 2 HOME PAGE 
 different design options

In order to text out what the design effect for internet  
use would have been for different types of design it
is necessary to set up a new design 
BUT all previous settings need to be cleared as
part of this process.
Unless this is done previous settings will remain
-------------------------------------------------------------*/
svyset [pweight=ind_wt],  clear( strata psu pweight )
svymean intuse , ci deff
svyset ,psu(psu)
svymean intuse , ci deff
svyset, clear(psu) strata(stratum)
svymean intuse , ci deff
svyset,psu(psu)
Now chi square tests for surveys
So the table with RC5 fails for same reasons 
s previously.

------------------------------------------------------------*/

svytab intuse sex, count row percent
svytab  intuse rc5,

svytab sex groc, count row percent format(%10.2f)
/*-------------------------------------------------------------

Now internet use by council area to illustrate 
different types of design effect for sub-groups

The srssubpop option in the second command gives
design effects compared to simple random sampling
in sub populations
-----------------------------------------------------------*/

 svymean intuse,  ci deff by(council) 
svymean intuse,  ci deff by(council) srssubpop

*--------------------------------------------------------
now some logistic regression
first get the dummy variables
-----------------------------------------------------------*/
tabulate groupinc, generate(groupinc)
tabulate groupinc1
/*----------------------------------
will miss out groupinc2 as analysis restricted to 
cases with income data
-----------------------------------------*/
svylogit intuse  groupinc3-groupinc6 if groupinc>0,prob deff deft or 
/*----------------------------------------------
compare with unweighted logistic regression
--------------------------------------------------*/
logistic intuse  groupinc3-groupinc6 if groupinc>0, or 

/*------------------------------------------------
now add urban rural you will see almost no effect
-----------------------------------------------------*/
tabulate shs_6cla, generate (rural)
svylogit intuse groupinc3-groupinc6 rural2-rural6 if groupinc>0,prob or

/*--------------------------------------------------
to fit a spline model you need to install the bsplines
package which is done easily by following the help
--------------------------------------------------------*/
 bspline,x(age)  power(3) gen(bs)

svymlogit intuse bs1-bs4,noconst deft deff
predict pr0 pr1
plot pr1 age