infile obs id age dist sex using "C:\Documents and Settings\Yijie\My Documents\ldata\dental.dat"
** save as .dta if you like;
drop obs
summ

** make in wide format, and save;

reshape wide dist, i(id) j(age)
reshape long dist, i(id) j(age)

** convert an ordinary data into a longitudinal dataset, specifying subject index and time index;
tsset id age


** Brief review on the LDA commands        ** ;
*********************************************;
** xtdes      Describe pattern of xt data;
** xtsum      Summarize xt data;
** xttab      Tabulate xt data;
** xtdata     Faster specification searches with xt data;
** xtreg      Fixed-, between- and random-effects, and population-averaged linear models;
** xtregar    Fixed- and random-effects linear models with an AR(1) disturbance;
** xtlogit    Fixed-effects, random-effects, & population-averaged logit models;
** xtpois     Fixed-effects, random-effects, & population-averaged Poisson models;
** xtgee      Population-averaged panel-data models using GEE;
********************************************************************;


** EDA analysis -- distance difference across boys and girls over time ** ;
** describe the pattern of data, including the missing pattern;
xtdes

sort age
by age: sum dist

graph dist, by(age) box
** in STATA 8, please do " graph box dist, by(age) ";**


** the time-varying variables are:  age, dist;
** the baseline variables are:  id,  sex;
** to summarize: means, standard deviations, frequenct, for time-series (xt) data;
xttab sex


**   Mean trend plot***;
xtgraph dist, group(sex) ti("Mean distance vs age")  bar(se)

** Spaghetti plots ** 
sort sex id age
graph dist age, by(sex) c(L) s(i)
** in STATA 8, please do  "twoway line dist age, by(sex) c(l) s(i)"; **


** kernel smooth **
ksm dist age, lowess gen(distsmth) nograph
graph dist distsmth age, c(.L.) s(.i.) 
** in STATA 8, please do  "twoway (scatter dist age) (connected distsmth age) "; **


gen distm = dist if sex == 0
gen distf = dist if sex == 1
ksm distm age, lowess gen(distmsmth)
graph dist distmsmth age, c(.L.) s(.io)
** in STATA 8, please do  "twoway (scatter dist age) (connected distmsmth age) "; **

ksm distf age, lowess gen(distfsmth)
graph dist distfsmth age, c(.L.) s(.io)
** in STATA 8, please do  "twoway (scatter dist age) (connected distfsmth age) "; **



** explore the within- and between subject variability **;
** For now, disregard the sex effect here;
** without adjusting age effects (time trend);
xtsumcorr dist

** The within-panel standard deviation is the same as that in{cmd:xtsum},
** except that the divisor of the variance is (N - n) instead of (N - 1), 
** where N is the total sample size and n is the number of subjects.
** The between-panel standard deviation is computed as the square root of the overall variance minus the within-panel variance.



** adjusting age effects (time trend);
** we need to use residual of Y after adjusting for age and sex effect;
xi: reg dist i.age sex
predict distres1, resid
xtsumcorr distres1


** last, to explore the correlation structure of response ***;
** first, to remove the effects of covariates, including age categories, sex;
** to calculate the autocorelation function, and plots **;
autocor distres1 age id

variogram distres1