能否在用egen时针对每个id直接只输出一个累计超额收益率？谢谢！

help collapse                                                                                                                           dialog:  collapse
---------------------------------------------------------------------------------------------------------------------------------------------------------

Title

    [D] collapse -- Make dataset of summary statistics


Syntax

        collapse clist [if] [in] [weight] [, options]

    where clist is either

        [(stat)] varlist [ [(stat)] ... ]

        [(stat)] target_var=varname [target_var=varname ...] [ [(stat)] ...]

    or any combination of the varlist or target_var forms, and stat is one of

        mean         means (default)
        median       medians
        p1           1st percentile
        p2           2nd percentile
        ...          3rd-49th percentiles
        p50          50th percentile (same as median)
        ...          51st-97th percentiles
        p98          98th percentile
        p99          99th percentile
        sd           standard deviations
        semean       standard error of the mean (sd/sqrt(n))
        sebinomial   standard error of the mean, binomial (sqrt(p(1-p)/n))
        sepoisson    standard error of the mean, Poisson (sqrt(mean))
        sum          sums
        rawsum       sums, ignoring optionally specified weight
        count        number of nonmissing observations
        max          maximums
        min          minimums
        iqr          interquartile range
        first        first value
        last         last value
        firstnm      first nonmissing value
        lastnm       last nonmissing value

    If stat is not specified, mean is assumed.

    options          description
    ---------------------------------------------------------------------------------------------------------------------------------------------------
    Options
      by(varlist)    groups over which stat is to be calculated
      cw             casewise deletion instead of all possible observations

    + fast           do not restore the original dataset should the user press Break; programmer's command
    ---------------------------------------------------------------------------------------------------------------------------------------------------
    + fast is not shown in the dialog box.
    varlist and varname in clist may contain time-series operators; see tsvarlist.
    aweights, fweights, iweights, and pweights are allowed; see weight, and see Weights below.  pweights may not be used with sd, semean, sebinomial,
      or sepoisson.  iweights may not be used with semean, sebinomial, or sepoisson.  aweights may not be used with sebinomial or sepoisson.


Menu

    Data > Create or change data > Other variable-transformation commands > Make dataset of means, medians, etc.


Description

    collapse converts the dataset in memory into a dataset of means, sums, medians, etc.  clist must refer to numeric variables exclusively.

    Note: See [D] contract if you want to collapse to a dataset of frequencies.


Options

        +---------+
    ----+ Options +------------------------------------------------------------------------------------------------------------------------------------

    by(varlist) specifies the groups over which the means, etc., are to be calculated.  If this option is not specified, the resulting dataset will
        contain 1 observation.  If it is specified, varlist may refer to either string or numeric variables.

    cw specifies casewise deletion.  If cw is not specified, all possible observations are used for each calculated statistic.

    The following option is available with collapse but is not shown in the dialog box:

    fast specifies that collapse not restore the original dataset should the user press Break.  fast is intended for use by programmers.


Weights

    collapse allows all four weight types; the default is aweights.  Weight normalization impacts only the sum, count, sd, semean, and sebinomial
    statistics.

    Here are the definitions for count and sum with weights:

     count:                           
        unweighted                    _N, the number of physical observations
        aweight:                      _N, the number of physical observations
        fweight, iweight, pweight:    sum(w_j), the sum of user-specified weights
     sum:                             
        unweighted                    sum(x_j), the sum of the variable
        aweight:                      sum(v_j*x_j); v_j = weights normalized to sum to _N
        fweight, iweight, pweight:    sum(w_j*x_j); w_j = user supplied weights.

    The sd statistic with weights returns the bias-corrected standard deviation, which is based on the factor sqrt(N/(N-1)), where N is the number of
    observations. Statistics sd, semean, sebinomial, and sepoisson are not allowed with pweighted data.  Otherwise, the statistic is changed by the
    weights through the computation of the count (N), as outlined above.

    For instance, consider a case in which there are 25 physical observations in the dataset and a weighting variable that sums to 57.  In the
    unweighted case, the weight is not specified, and N = 25.  In the analytically weighted case, N is still 25; the scale of the weight is irrelevant.
    In the frequency-weighted case, however, N = 57, the sum of the weights.

    The rawsum statistic with aweights ignores the weight, with one exception:  observations with zero weight will not be included in the sum.


Examples

    -----------------------------------------------------------------------------------------------------------------------------------------------------
    Setup
        . webuse college
        . describe
        . list

    Create dataset containing the 25th percentile of gpa for each year
        . collapse (p25) gpa [fw=number], by(year)

    List the result
        . list

    -----------------------------------------------------------------------------------------------------------------------------------------------------
    Setup
        . webuse college, clear

    Create dataset containing the mean and median of gpa and hour for each year, and store median of gpa and hour in medgpa and medhour, respectively
        . collapse (mean) gpa hour (median) medgpa=gpa medhour=hour [fw=number], by(year)

    List the result
        . list

    -----------------------------------------------------------------------------------------------------------------------------------------------------
    Setup
        . webuse college, clear

    Create dataset containing the count of gpa and hour and the minimums of gpa and hour, and store the minimums in mingpa and minhour, respectively
        . collapse (count) gpa hour (min) mingpa=gpa minhour=hour [fw=number], by(year)

    List the result
        . list

    -----------------------------------------------------------------------------------------------------------------------------------------------------
    Setup
        . webuse college, clear
        . replace gpa = . in 2/4

    Create dataset containing the mean of gpa and hour for each year, but ignore all observations that have missing values when calculating the means
        . collapse (mean) gpa hour [fw=number], by(year) cw

    List the result
        . list
    -----------------------------------------------------------------------------------------------------------------------------------------------------


Also see

    Manual:  [D] collapse

      Help:  [D] contract, [D] egen, [D] statsby, [R] summarize

by id, sort: drop if car == car[_n+1]

谢谢！
发现用这个就不必剔除样本，这样以便于进一步计算不同窗口期的收益率：
by car,sort: replace car =. if car== car[_n+1]

楼主你好，我最近也要用事件研究法，可是因为没找到系统的文献或者书籍，我本身对事件研究法也是第一次接触，所以能不能请你给我一些指点，比如说，能不能给我介绍点文章或者资料。万分感谢！