cd c:\archive\fac_council
set more off
capture log close
log using tables.log, replace

/* -----------------------------------------------------------
* Table of means and mean differences and cohen's d
* using data for faculty/course match across the two semesters
-------------------------------------------------------------- */
use temp_for_analysis.dta, clear
keep if match==1 
drop match
format q* %9.3f
drop  q1_1_10 - q4_2_10   q1_1_09 -  q4_5_09
drop q*_*_*
drop sd_q*_*_*
drop rate*
list group n* *q5* in 1/10, noobs clean
gen n_d = n_10+n_09
gen n_sec_d = n_sec_10 + n_sec_09

forvalues i=5/13  {
	disp "`i'"
	count if sd_q`i'_09 == 0
	count if sd_q`i'_10 == 0
	count if sd_q`i'_10 == 0 & sd_q`i'_09 == 0
}

list n_10 n_09 if sd_q7_10 == 0 & sd_q7_09 == 0

forvalues i=5/13  {
	gen q`i'_d = q`i'_10 - q`i'_09
}

forvalues i=5/13  {
	gen c_q`i'_d = q`i'_d /sd_q`i'_09
}

drop sd_*

reshape long q5 q6 q7 q8 q9 q10 q11 q12 q13 n n_sec ///
	c_q5 c_q6 c_q7 c_q8 c_q9 c_q10 c_q11 c_q12 c_q13, i(group) j(compare) str

sort compare group
list compare group n n_sec q5 c_q5 in 1/10
list compare group n n_sec q5 c_q5 in 414/423
reshape long q c_q, i(compare group) j(item)
list compare group item n n_sec q c_q in 1/10
list if  compare=="_d" & group==306
count if c_q==. & compare=="_d"

table item if  compare=="_d" [aweight=n] , c( min q p25 q mean q p75 q max q ) format(%9.3f)
table item if  compare=="_d" [aweight=n] , c( min c_q p25 q mean c_q p75 c_q max c_q ) format(%9.3f)


statsby mean = r(mean) se = r(se) nobs =r(N) ub = r(ub) lb= r(lb), ///
	by(compare item ) saving(matched_overall.dta, replace) : ci q [aweight=n] 

statsby mean = r(mean) , ///
	by(compare item ) saving(c_matched_overall.dta, replace) : ci c_q [aweight=n] 

statsby n_secs = r(sum) , ///
	by(compare item) saving(matched_overall_n_sec.dta, replace) : summ n_sec 
	
statsby n = r(sum) , ///
	by(compare item) saving(matched_overall_n.dta, replace) : summ n
	
statsby mean = r(mean) se = r(se) nobs =r(N) ub = r(ub) lb= r(lb), ///
	by(ctype compare item) saving(matched_by_ctypes.dta, replace) : ci q [aweight=n]

statsby mean = r(mean) , ///
	by(ctype compare item) saving(c_matched_by_ctypes.dta, replace) : ci c_q [aweight=n]
	
statsby n_secs = r(sum) , ///
	by(ctype compare item) saving(matched_by_ctypes_n_sec.dta, replace) : summ n_sec 
	
statsby n = r(sum) , ///
	by(ctype compare item) saving(matched_by_ctypes_n.dta, replace) : summ n 


* Table for Overall
use c_matched_overall.dta, clear
rename mean cd
merge 1:1 compare item using matched_overall.dta
assert _merge==3
drop _merge
merge 1:1 compare item using matched_overall_n_sec.dta
assert _merge==3
drop _merge
merge 1:1 compare item using matched_overall_n.dta
assert _merge==3
drop _merge

format mean se ub lb cd %9.2f
gen p = 2*ttail(nobs,abs(mean/se)) if compare=="_d"

format p %9.3f
tostring p, force usedisplayformat gen(p_value)
replace p_value = "<0.001" if p<0.001
drop p

gen sd = se*sqrt(nobs) 
format sd %9.2f


foreach var of varlist mean sd se  ub lb cd {
	tostring `var', replace force usedisplayformat
}

gen ci  = "(" + lb + ", " + ub + ")"	
list, sepby(compare)
drop se ub lb

reshape wide nobs p_value mean  ci n_secs n sd cd , i(item) j(compare) str
drop p_value_10 p_value_09 ci_09 ci_10 

tostring item nobs* n_secs* mean* cd_d, force replace
tostring n_09 n_10 n_d, force replace format(%9.0fc)

list item 

set obs 12
replace item = "Observations" if _n== 10
replace item = "Sections" if _n == 11
replace item = "Responses"  if _n ==12
foreach x in _09 _10 {
	replace sd`x' = nobs`x'[1] if _n == 10
	replace sd`x' = n_secs`x'[1] if _n == 11
	replace sd`x' = n`x'[1] if _n ==12
}



list item mean_09 sd_09 mean_10 sd_10 mean_d sd_d ci_d p_value_d cd_d, noobs clean
list n* in 1/1
drop n*
listtex item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d cd_d , type



* tables by course types.
use c_matched_by_ctypes.dta, clear
rename mean cd
merge 1:1 ctype compare item using matched_by_ctypes.dta
assert _merge ==3
drop _merge
merge 1:1 ctype compare item using matched_by_ctypes_n_sec.dta
assert _merge==3
drop _merge
merge 1:1 ctype compare item using matched_by_ctypes_n.dta
assert _merge==3
drop _merge

format mean se ub lb cd %9.2f
gen p = 2*ttail(nobs,abs(mean/se)) if compare=="_d"


format p %9.3f
tostring p, force usedisplayformat gen(p_value)
tostring p, force format(%9.2f) gen(p2)
replace p_value = p2 if p>0.01
replace p_value = "<0.001" if p<0.001
drop p p2

gen sd = se*sqrt(nobs) 
format sd %9.2f


foreach var of varlist mean sd se  ub lb cd {
	tostring `var', replace force usedisplayformat
}

gen ci  = "(" + lb + ", " + ub + ")"	
list, sepby(compare)
drop se ub lb


reshape wide nobs p_value mean ci n_secs n sd cd, i(ctype item) j(compare) str
drop p_value_10 p_value_09 ci_09 ci_10 

tostring item nobs* n_secs* mean* cd_d, force replace
tostring n_09 n_10 n_d, force replace format(%9.0fc)


levelsof ctype, local(levels)
foreach l of local levels {
	preserve
	keep if ctype=="`l'"
	assert _N==9
	set obs 12
	replace item = "Observations" if _n== 10
	replace item = "Sections" if _n == 11
	replace item = "Responses"  if _n ==12
	foreach x in _09 _10  {
		replace sd`x' = nobs`x'[1] if _n == 10
		replace sd`x' = n_secs`x'[1] if _n == 11
		replace sd`x' = n`x'[1] if _n ==12
	}
	drop n*
	disp "*******************************************************"
	disp "`l'"
	list item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d p_value_d cd_d, noobs clean
	disp "              "
	listtex item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d cd_d , type
	disp "*******************************************************"
	restore
}


/* -----------------------------------------------------------
* Table of means and mean differences
* Note this code gathers up means for *all* in 09 and 10, but 
* mean differences only for "matched" groups -- tables later modified
* to show only summaries for 09 and 10.
-------------------------------------------------------------- */
use temp_for_analysis.dta, clear
drop sd_*
format q* %9.3f
drop  q1_1_10 - q4_2_10   q1_1_09 -  q4_5_09
drop q*_*_*
drop rate*
list group n* q5* in 1/10, noobs clean
gen n_d = n_10+n_09
gen n_sec_d = n_sec_10 + n_sec_09

forvalues i=5/13  {
	gen q`i'_d = q`i'_10 - q`i'_09
}

reshape long q5 q6 q7 q8 q9 q10 q11 q12 q13 n n_sec , i(group) j(compare) str

sort compare group
list compare group n n_sec q5 in 1/10
drop if n==. & q5==.
reshape long q, i(compare group) j(item)
list compare group item n n_sec q in 1/10

list if  compare=="_d" & group==306

table item if  compare=="_d" [aweight=n] , c( min q p25 q mean q p75 q max q ) format(%9.3f)

statsby mean = r(mean) se = r(se) nobs =r(N) ub = r(ub) lb= r(lb), ///
	by(compare item ) saving(overall.dta, replace) : ci q [aweight=n] 

statsby n_secs = r(sum) , ///
	by(compare item) saving(overall_n_sec.dta, replace) : summ n_sec 
	
statsby n = r(sum) , ///
	by(compare item) saving(overall_n.dta, replace) : summ n
	
statsby mean = r(mean) se = r(se) nobs =r(N) ub = r(ub) lb= r(lb), ///
	by(ctype compare item) saving(by_ctypes.dta, replace) : ci q [aweight=n]
	
statsby n_secs = r(sum) , ///
	by(ctype compare item) saving(by_ctypes_n_sec.dta, replace) : summ n_sec 
	
statsby n = r(sum) , ///
	by(ctype compare item) saving(by_ctypes_n.dta, replace) : summ n 


* Table for Overall
use overall.dta, clear
merge 1:1 compare item using overall_n_sec.dta
assert _merge==3
drop _merge
merge 1:1 compare item using overall_n.dta
assert _merge==3
drop _merge

format mean se ub lb %9.2f
gen p = 2*ttail(nobs,abs(mean/se)) if compare=="_d"

format p %9.3f
tostring p, force usedisplayformat gen(p_value)
replace p_value = "<0.001" if p<0.001
drop p

gen sd = se*sqrt(nobs) 
format sd %9.2f


foreach var of varlist mean sd se  ub lb {
	tostring `var', replace force usedisplayformat
}

gen ci  = "(" + lb + ", " + ub + ")"	
list, sepby(compare)
drop se ub lb

reshape wide nobs p_value mean ci n_secs n sd , i(item) j(compare) str
drop p_value_10 p_value_09 ci_09 ci_10 

tostring item nobs* n_secs* mean* , force replace
tostring n_09 n_10 n_d, force replace format(%9.0fc)

list item 

set obs 12
replace item = "Observations" if _n== 10
replace item = "Sections" if _n == 11
replace item = "Responses"  if _n ==12
foreach x in _09 _10 _d {
	replace sd`x' = nobs`x'[1] if _n == 10
	replace sd`x' = n_secs`x'[1] if _n == 11
	replace sd`x' = n`x'[1] if _n ==12
}



list item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d p_value_d , noobs clean
list n* in 1/1
drop n*
listtex item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d , type



* tables by course types.
use by_ctypes.dta, clear
merge 1:1 ctype compare item using by_ctypes_n_sec.dta
assert _merge==3
drop _merge
merge 1:1 ctype compare item using by_ctypes_n.dta
assert _merge==3
drop _merge

format mean se ub lb %9.2f
gen p = 2*ttail(nobs,abs(mean/se)) if compare=="_d"


format p %9.3f
tostring p, force usedisplayformat gen(p_value)
tostring p, force format(%9.2f) gen(p2)
replace p_value = p2 if p>0.01
replace p_value = "<0.001" if p<0.001
drop p p2

gen sd = se*sqrt(nobs) 
format sd %9.2f


foreach var of varlist mean sd se  ub lb {
	tostring `var', replace force usedisplayformat
}

gen ci  = "(" + lb + ", " + ub + ")"	
list, sepby(compare)
drop se ub lb


reshape wide nobs p_value mean ci n_secs n sd , i(ctype item) j(compare) str
drop p_value_10 p_value_09 ci_09 ci_10 

tostring item nobs* n_secs* mean* , force replace
tostring n_09 n_10 n_d, force replace format(%9.0fc)


levelsof ctype, local(levels)
foreach l of local levels {
	preserve
	keep if ctype=="`l'"
	assert _N==9
	set obs 12
	replace item = "Observations" if _n== 10
	replace item = "Sections" if _n == 11
	replace item = "Responses"  if _n ==12
	foreach x in _09 _10 _d {
		replace sd`x' = nobs`x'[1] if _n == 10
		replace sd`x' = n_secs`x'[1] if _n == 11
		replace sd`x' = n`x'[1] if _n ==12
	}
	drop n*
	disp "*******************************************************"
	disp "`l'"
	list item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d p_value_d , noobs clean
	disp "              "
	listtex item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d , type
	disp "*******************************************************"
	restore
}


/* -----------------------------------------------------------
* proportions -- only for "matched" faculty/course combinations
-------------------------------------------------------------- */
set more off
use temp_for_analysis.dta, clear
keep if match==1
format q* %9.3f
drop  q1_1_10- q4_2_10   q1_1_09 -  q4_5_09
drop   q5_10- q13_10  q5_09- q13_09
drop rate*
drop sd_*
list group n* q5* in 1/10, noobs clean
gen n_d = n_10+n_09
gen n_sec_d = n_sec_10 + n_sec_09

forvalues i=5/13  {
	forvalues j=1/5 {
		gen q`i'_`j'_d = q`i'_`j'_10 - q`i'_`j'_09
	}
}

reshape long n n_sec ///
	q5_1 q5_2 q5_3 q5_4 q5_5 ///
	q6_1 q6_2 q6_3 q6_4 q6_5 ///
	q7_1 q7_2 q7_3 q7_4 q7_5 ///
	q8_1 q8_2 q8_3 q8_4 q8_5 ///
	q9_1 q9_2 q9_3 q9_4 q9_5 ///
	q10_1 q10_2 q10_3 q10_4 q10_5 ///
	q11_1 q11_2 q11_3 q11_4 q11_5 ///
	q12_1 q12_2 q12_3 q12_4 q12_5 ///
	q13_1 q13_2 q13_3 q13_4 q13_5, ///
	i(group) j(compare) str
	
list group compare n n_sec q5* in 1/10 

reshape long q5 q6 q7 q8 q9 q10 q11 q12 q13 , i(compare group) j(rating) str
sort compare group
list compare group n n_sec rating q5 in 1/20
drop if n==. & q5==.
reshape long q, i(compare group rating) j(item)
sort compare group item rating 
list compare group item n n_sec rating q in 1/10

list compare group rating item match q if compare=="_d" & group==306 , sepby(item)
table item if rating=="_5" & compare=="_d" [aweight=n] , c( min q p25 q mean q p75 q max q ) format(%9.2f)
table item if rating=="_1" & compare=="_d" [aweight=n] , c( min q p25 q mean q p75 q max q ) format(%9.2f)

statsby mean = r(mean) se = r(se) nobs =r(N) ub = r(ub) lb= r(lb), ///
	by(compare item rating) saving(matched_prop.dta, replace) : ci q [aweight=n] 

statsby n_secs = r(sum) , ///
	by(compare item rating) saving(matched_prop_n_sec.dta, replace) : summ n_sec 
	
statsby n = r(sum) , ///
	by(compare item rating) saving(matched_prop_n.dta, replace) : summ n
	
statsby mean = r(mean) se = r(se) nobs =r(N) ub = r(ub) lb= r(lb), ///
	by(ctype compare item rating ) saving(matched_ctypes_prop.dta, replace) : ci q [aweight=n]
	
statsby n_secs = r(sum) , ///
	by(ctype compare item rating ) saving(matched_ctypes_prop_n_sec.dta, replace) : summ n_sec 
	
statsby n = r(sum) , ///
	by(ctype compare item rating ) saving(matched_ctypes_prop_n.dta, replace) : summ n
	

use matched_prop.dta, clear
merge 1:1 compare item rating using matched_prop_n_sec.dta
assert _merge ==3
drop _merge
merge 1:1 compare item rating using matched_prop_n.dta
assert _merge ==3
drop _merge

list compare item rating n* mean in 1/10

replace rating = substr(rating,2,1)
destring rating, replace

gen sd = se*sqrt(nobs) if rating > 1
format mean se sd ub lb %9.3f

gen p = 2*ttail(nobs,abs(mean/se)) if compare=="_d" & rating>1

format p %9.3f
tostring p, force usedisplayformat gen(p_value)
tostring p, force format(%9.2f) gen(p2)
replace p_value = p2 if p>0.01
replace p_value = "<0.001" if p<0.001
drop p p2


foreach var of varlist mean sd se  ub lb {
	tostring `var', replace force usedisplayformat
}

gen ci  = "(" + lb + ", " + ub + ")" if rating >1 & mean != "0.000"
list in 126/135 
drop se ub lb

reshape wide nobs p_value mean ci n_secs n sd , i(item rating) j(compare) str
drop p_value_10 p_value_09 ci_09 ci_10 

gsort +item -rating
list item rating *_d in 1/10
replace item = . if rating < 5

tostring item rating nobs* n_secs* mean* , force replace
tostring n_09 n_10 n_d, force replace format(%9.0fc)

foreach x in item rating sd_09 sd_10 sd_d p_value_d {
	replace `x' = "" if `x'=="."
}

count
set obs 48
replace item = "Observations" if _n== 46
replace item = "Sections" if _n == 47
replace item = "Responses"  if _n ==48
foreach x in _09 _10  {
	replace sd`x' = nobs`x'[1] if _n == 46
	replace sd`x' = n_secs`x'[1] if _n == 47
	replace sd`x' = n`x'[1] if _n == 48
}


list item rating mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d p_value_d , noobs clean
list n* in 1/1
drop n*
listtex item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d , type


* by course type
use matched_ctypes_prop.dta, clear
merge 1:1 compare ctype item rating using matched_ctypes_prop_n_sec.dta
assert _merge ==3
drop _merge
merge 1:1 compare ctype item rating using matched_ctypes_prop_n.dta
assert _merge ==3
drop _merge


list compare item rating ctype n* mean in 1/10

replace rating = substr(rating,2,1)
destring rating, replace

gen sd = se*sqrt(nobs) if rating > 1
format mean se sd ub lb %9.3f

gen p = 2*ttail(nobs,abs(mean/se)) if compare=="_d" & rating>1

format p %9.3f
tostring p, force usedisplayformat gen(p_value)
tostring p, force format(%9.2f) gen(p2)
replace p_value = p2 if p>0.01
replace p_value = "<0.001" if p<0.001
drop p p2


foreach var of varlist mean sd se  ub lb {
	tostring `var', replace force usedisplayformat
}

gen ci  = "(" + lb + ", " + ub + ")" if rating >1
list in 801/810 , noobs clean
drop se ub lb

reshape wide nobs p_value mean ci n_secs n sd , i(ctype item rating) j(compare) str
drop p_value_10 p_value_09 ci_09 ci_10 


gsort ctype +item -rating
list ctype item rating *_d in 1/10
replace item = . if rating < 5

tostring item rating nobs* n_secs* mean* , force replace
tostring n_09 n_10 n_d, force replace format(%9.0fc)

foreach x in item rating sd_09 sd_10 sd_d p_value_d {
	replace `x' = "" if `x'=="."
}

levelsof ctype, local(levels)
foreach l of local levels {
	preserve
	keep if ctype =="`l'"
	assert _N == 45
	set obs 48
	replace item = "Observations" if _n== 46
	replace item = "Sections" if _n == 47
	replace item = "Responses"  if _n ==48
	foreach x in _09 _10 {
		replace sd`x' = nobs`x'[1] if _n == 46
		replace sd`x' = n_secs`x'[1] if _n == 47
		replace sd`x' = n`x'[1] if _n == 48
	}

	disp "**********************************************************************"
	disp "`l'"
	list item rating mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d p_value_d , noobs clean
	list n* in 1/1
	drop n*
	listtex item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d , type

	restore
}







/* -----------------------------------------------------------
* proportions
* Note this code gathers up means for *all* in 09 and 10, but 
* mean differences only for "matched" groups -- tables later modified
* to show only summaries for 09 and 10.
-------------------------------------------------------------- */
set more off
use temp_for_analysis.dta, clear
drop sd_*
format q* %9.3f
drop  q1_1_10- q4_2_10   q1_1_09 -  q4_5_09
drop   q5_10- q13_10  q5_09- q13_09
drop rate*
list group n* q5* in 1/10, noobs clean
gen n_d = n_10+n_09
gen n_sec_d = n_sec_10 + n_sec_09

forvalues i=5/13  {
	forvalues j=1/5 {
		gen q`i'_`j'_d = q`i'_`j'_10 - q`i'_`j'_09
	}
}

reshape long n n_sec ///
	q5_1 q5_2 q5_3 q5_4 q5_5 ///
	q6_1 q6_2 q6_3 q6_4 q6_5 ///
	q7_1 q7_2 q7_3 q7_4 q7_5 ///
	q8_1 q8_2 q8_3 q8_4 q8_5 ///
	q9_1 q9_2 q9_3 q9_4 q9_5 ///
	q10_1 q10_2 q10_3 q10_4 q10_5 ///
	q11_1 q11_2 q11_3 q11_4 q11_5 ///
	q12_1 q12_2 q12_3 q12_4 q12_5 ///
	q13_1 q13_2 q13_3 q13_4 q13_5, ///
	i(group) j(compare) str
	
list group compare n n_sec q5* in 1/10 

reshape long q5 q6 q7 q8 q9 q10 q11 q12 q13 , i(compare group) j(rating) str
sort compare group
list compare group n n_sec rating q5 in 1/20
drop if n==. & q5==.
reshape long q, i(compare group rating) j(item)
sort compare group item rating 
list compare group item n n_sec rating q in 1/10

list compare group rating item match q if compare=="_d" & group==306 , sepby(item)
table item if rating=="_5" & compare=="_d" [aweight=n] , c( min q p25 q mean q p75 q max q ) format(%9.2f)
table item if rating=="_1" & compare=="_d" [aweight=n] , c( min q p25 q mean q p75 q max q ) format(%9.2f)

statsby mean = r(mean) se = r(se) nobs =r(N) ub = r(ub) lb= r(lb), ///
	by(compare item rating) saving(prop.dta, replace) : ci q [aweight=n] 

statsby n_secs = r(sum) , ///
	by(compare item rating) saving(prop_n_sec.dta, replace) : summ n_sec 
	
statsby n = r(sum) , ///
	by(compare item rating) saving(prop_n.dta, replace) : summ n
	
statsby mean = r(mean) se = r(se) nobs =r(N) ub = r(ub) lb= r(lb), ///
	by(ctype compare item rating ) saving(ctypes_prop.dta, replace) : ci q [aweight=n]
	
statsby n_secs = r(sum) , ///
	by(ctype compare item rating ) saving(ctypes_prop_n_sec.dta, replace) : summ n_sec 
	
statsby n = r(sum) , ///
	by(ctype compare item rating ) saving(ctypes_prop_n.dta, replace) : summ n
	

use prop.dta, clear
merge 1:1 compare item rating using prop_n_sec.dta
assert _merge ==3
drop _merge
merge 1:1 compare item rating using prop_n.dta
assert _merge ==3
drop _merge

list compare item rating n* mean in 1/10

replace rating = substr(rating,2,1)
destring rating, replace

gen sd = se*sqrt(nobs) if rating > 1
format mean se sd ub lb %9.3f

gen p = 2*ttail(nobs,abs(mean/se)) if compare=="_d" & rating>1

format p %9.3f
tostring p, force usedisplayformat gen(p_value)
tostring p, force format(%9.2f) gen(p2)
replace p_value = p2 if p>0.01
replace p_value = "<0.001" if p<0.001
drop p p2


foreach var of varlist mean sd se  ub lb {
	tostring `var', replace force usedisplayformat
}

gen ci  = "(" + lb + ", " + ub + ")" if rating >1 & mean != "0.000"
list in 126/135 
drop se ub lb

reshape wide nobs p_value mean ci n_secs n sd , i(item rating) j(compare) str
drop p_value_10 p_value_09 ci_09 ci_10 

gsort +item -rating
list item rating *_d in 1/10
replace item = . if rating < 5

tostring item rating nobs* n_secs* mean* , force replace
tostring n_09 n_10 n_d, force replace format(%9.0fc)

foreach x in item rating sd_09 sd_10 sd_d p_value_d {
	replace `x' = "" if `x'=="."
}

count
set obs 48
replace item = "Observations" if _n== 46
replace item = "Sections" if _n == 47
replace item = "Responses"  if _n ==48
foreach x in _09 _10 _d {
	replace sd`x' = nobs`x'[1] if _n == 46
	replace sd`x' = n_secs`x'[1] if _n == 47
	replace sd`x' = n`x'[1] if _n == 48
}


list item rating mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d p_value_d , noobs clean
list n* in 1/1
drop n*
listtex item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d , type


* by course type
use ctypes_prop.dta, clear
merge 1:1 compare ctype item rating using ctypes_prop_n_sec.dta
assert _merge ==3
drop _merge
merge 1:1 compare ctype item rating using ctypes_prop_n.dta
assert _merge ==3
drop _merge


list compare item rating ctype n* mean in 1/10

replace rating = substr(rating,2,1)
destring rating, replace

gen sd = se*sqrt(nobs) if rating > 1
format mean se sd ub lb %9.3f

gen p = 2*ttail(nobs,abs(mean/se)) if compare=="_d" & rating>1

format p %9.3f
tostring p, force usedisplayformat gen(p_value)
tostring p, force format(%9.2f) gen(p2)
replace p_value = p2 if p>0.01
replace p_value = "<0.001" if p<0.001
drop p p2


foreach var of varlist mean sd se  ub lb {
	tostring `var', replace force usedisplayformat
}

gen ci  = "(" + lb + ", " + ub + ")" if rating >1
list in 801/810 , noobs clean
drop se ub lb

reshape wide nobs p_value mean ci n_secs n sd , i(ctype item rating) j(compare) str
drop p_value_10 p_value_09 ci_09 ci_10 


gsort ctype +item -rating
list ctype item rating *_d in 1/10
replace item = . if rating < 5

tostring item rating nobs* n_secs* mean* , force replace
tostring n_09 n_10 n_d, force replace format(%9.0fc)

foreach x in item rating sd_09 sd_10 sd_d p_value_d {
	replace `x' = "" if `x'=="."
}

levelsof ctype, local(levels)
foreach l of local levels {
	preserve
	keep if ctype =="`l'"
	assert _N == 45
	set obs 48
	replace item = "Observations" if _n== 46
	replace item = "Sections" if _n == 47
	replace item = "Responses"  if _n ==48
	foreach x in _09 _10 _d {
		replace sd`x' = nobs`x'[1] if _n == 46
		replace sd`x' = n_secs`x'[1] if _n == 47
		replace sd`x' = n`x'[1] if _n == 48
	}

	disp "**********************************************************************"
	disp "`l'"
	list item rating mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d p_value_d , noobs clean
	list n* in 1/1
	drop n*
	listtex item mean_09 sd_09  mean_10 sd_10 mean_d sd_d ci_d , type

	restore
}


/*-----------------------------------------------------------
* Scatterplots and univariate non-parametric regressions;
* change in rating and change in response rate
* change in rating and 2009 response rate
-------------------------------------------------------------*/
cd C:\archive\fac_council
use temp_for_analysis.dta, clear
keep if match==1 
drop match
format q* %9.3f
drop  q1_1_10 - q4_2_10   q1_1_09 -  q4_5_09
drop q*_*_*
drop sd_*
list group n* *q5* in 1/10, noobs clean
gen n_d = n_10+n_09
gen n_sec_d = n_sec_10 + n_sec_09


forvalues i=5/13  {
	gen q`i'_d = q`i'_10 - q`i'_09
}


gen rate_d = rate_10 - rate_09

format rate_d %9.2f
format q*_d %9.3f

list *rate* *q5* ctype grp_last in 1/10, clean 


forvalues i=5/13 {
	lpoly q`i'_d rate_d [aweight=n_d], scheme(lean2) ci title("") note("") noscatter ///
		legend(off) ytitle("Difference in Rating") xtitle("Difference in Response Rate") ///
		ylabel(-2(1)1) addplot( scatter q`i'_d rate_d [aweight=n_d] , msymbol(circle_hollow) ) ///
		saving(rate_d_q`i', replace)
	lpoly q`i'_d rate_09 [aweight=n_d], scheme(lean2) ci title("") note("") noscatter ///
		legend(off) ytitle("Difference in Rating") xtitle("2009 Response Rate") ///
		ylabel(-2(1)1) addplot( scatter q`i'_d rate_09 [aweight=n_d] , msymbol(circle_hollow) ) ///
		saving(rate_09_q`i', replace)
	graph combine rate_d_q`i'.gph  rate_09_q`i'.gph , ///
		col(1) scheme(lean2) ysize(7) iscale(0.7)
	graph export testq`i'.emf, replace
}



/* ---------------------------------------------------------------------
* detailed summary statistics for appendix  -- percentiles of mean
------------------------------------------------------------------------ */
set more off
use temp_for_analysis.dta, clear
drop sd_*
format q* %9.2f
drop  q1_1_10- q13_5_10  q1_1_09- q13_5_09
drop match grp_last
order ctype , after(group)
list group *n* q5_10 q5_09 in 1/10, noobs clean
reshape long n n_sec rate q5 q6 q7 q8 q9 q10 q11 q12 q13 , i(group) j(t) str
list group - q5 in 1/10
rename rate q14
reshape long q , i(group t) j(item)
drop if q==.
gen year = 2009 if t =="_09"
replace year = 2010 if t=="_10"
drop t
list in 1/10

collapse (p5) p5=q  (p10) p10=q  (p25) p25=q  ///
	(p50) p50=q  (p75) p75=q  ///
	(p90) p90=q  (p95) p95=q (rawsum) sum=n sum_sec = n_sec  [aweight=n], by(item year)

list year item p5 p10 p25 p50 p75 p90 p95 sum*
reshape wide p5-sum_sec , i(item) j(year)

sort item
tostring sum*, replace force format(%9.0fc)
tostring p*, replace force format(%9.2f)
tostring item, replace
replace item="Response Rate (%)" if item=="14"
list item p50* sum* 
set obs 12
replace item = "Sections" if _n==11
replace p502009 = sum_sec2009[1] if _n==11
replace p502010 = sum_sec2010[1] if _n==11
replace item = "Responses" if _n==12
replace p502009 = sum2009[1] if _n==12
replace p502010 = sum2010[1] if _n==12
list item p50* sum* 
drop sum*

list item *2009, noobs clean
list item *2010, noobs clean
listtex item *2009 *2010, type

* course types
set more off
use temp_for_analysis.dta, clear
drop sd_*
format q* %9.3f
drop  q1_1_10- q13_5_10  q1_1_09- q13_5_09
drop match grp_last
order ctype , after(group)
list group *n* q5_10 q5_09 in 1/10, noobs clean
reshape long n n_sec rate q5 q6 q7 q8 q9 q10 q11 q12 q13 , i(group) j(t) str
list in 1/10
rename rate q14
reshape long q , i(group t) j(item)
drop if q==.
gen year = 2009 if t =="_09"
replace year = 2010 if t=="_10"
drop t

levelsof ctype, local(levels)
foreach l of local levels {
	preserve
	keep if ctype=="`l'"	

	collapse (p5) p5=q  (p10) p10=q  (p25) p25=q  ///
		(p50) p50=q  (p75) p75=q  ///
		(p90) p90=q  (p95) p95=q (rawsum) sum=n sum_sec = n_sec  [aweight=n], by(item year)

	reshape wide p5-sum_sec , i(item) j(year)
	sort item
	tostring sum*, replace force format(%9.0fc)
	tostring p*, replace force format(%9.2f)
	tostring item, replace
	replace item="Response Rate (%)" if item=="14"
	set obs 12
	replace item = "Sections" if _n==11
	replace p502009 = sum_sec2009[1] if _n==11
	replace p502010 = sum_sec2010[1] if _n==11
	replace item = "Responses" if _n==12
	replace p502009 = sum2009[1] if _n==12
	replace p502010 = sum2010[1] if _n==12
	list item p50* sum* 
	drop sum*
	disp "`l'"
	list item *2009, noobs clean
	list item *2010, noobs clean
	listtex item *2009 *2010, type
	restore
}

log close