.- help for ^sumby^ > Version: 3/4/01 .-^Summary statistics in tabular format^ ^------------------------------------^

^sumby^ varlist [^if^ exp] [^in^ range] [^,^ ^b^y(var2) ^st^at^(^statistics^) > ^

^l^abel^(^labels^)^ ^d^ec^(^numbers^)^ ^w^idth^(^#^)^ ^h^ead ^m^is no^B > YT^ot]

^start(^int 12^)^ ^te^st^(^string^)^

^Description^ ^-----------^

Generates a table with summary statistics of variable ^var1^. If ^by(var2)^ is specified, the statistics are calculated for each variabel in the ^varlist^ and for each value of var2, and displayed in columns in order of values of var2. A "total"-column with the values of the statistics of var1 for all (non-missing) values of var2 is displayed, unless ^nobytot^ is specified.

^Options^ ^-------^

^by(^var2^)^ Optional. If not specified only overall summary statistics

are calculated.

^stat(^st..^)^ Optional, specifies the requested statistics of var1. Default is 4 statistics: mean;sd median range number. ^st..^ Lowercase and UPPERCASE indifferent! Separators:"," or " ". ^----^ Specifies Default label

^mean^,^m^ mean Mean ^median^,^med^,^q2^ median Median ^sd^ standard deviation SD ^msd^ mean;sd Mean;SD ^min^ minimum Minimum ^max^ maximum Maximum ^range^,^r^ min-max Range ^q1^ Q1 Q1 ^q3^ Q3 Q3 ^p01^,^p1^ 1th percentile 1 % ^p5^,^p05^ 5th percentile 5 % ^p10^ 10th percentile 10 % ^p90^ 90th percentile 90 % ^p95^ 95th percentile 95 % ^p99^ 99th percentile 99 % ^n^ number of non missing N

^all^ or ^_all^ is equivalent to "n msd range p5 q1 median q3 p95" If "all" appears anywhere in the string, requests for other statistics will be ignored.

^label(^..^)^ Specifies labels in the order of requested statistics. Strings are separated by " " or ",". A "." defines the default label for the corresponding statistic. The default label is also taken if not specified.

^dec(^..^)^ Specifies number of decimals per statistic in the order specified by ^stat()^. The default value is, for each statistic, the number of decimals specified in the format attached to the variable var1 (between 0 and 9). The numbers should be again comma or blank separated. A "." instead of a number indicates the default format. Frequencies are always displayed without decimals. If the length of the statistic would exceed the available width for the display, the number of decimals is reduced.

^width^ Integer number, with default value 10. Specifies the width for each column with statistics. The value 10 corresponds to the width of columns created by tabulate. If the length of the string required to display a statistic exceeds this width, the width of all the columns is increased. The width excludes a space between columns. Long columnlabels in the header are abbreviated if necessary.

^mis^ Specifies that the category with missing values of ^var2^ should be included in the calculations. Only applicable if ^by(var2)^ is specified. Default: observations with missing value on var2 are excluded.

^head^ Request the display of a heading of the table. Active only if ^by(var2)^ is specified.

^nobytot^ Request suppression of a total column if ^by()^ is specified.

^start()^ Specifies starting point of the body of the table, i.e. the column in which the ^|^ is printed. Default 12. Values less than 12, will be replaced by 12.

^test(XXX)^ request the calculation of a P-value: with XXX = ^kw^, ^k-w^ or ^wil^coxon or ^rank^sum: Kruskal-Wallis test XXX = np^tr^end: NPtrend test by ^by^-variable ^(*)^ XXX = ^sp^ear: Spearman rank correlation test by ^by^-variable ^(*)^ ^(*)^ the P-value calculated by ^NPtrend^ depends on the values of the ^by^-va > riable: ^by^ = 1/2/3, ^by^ = 2/3/4 and ^by^ = 2/5/8 provide the same P-values, but > they may be different when ^by^ = 1/2/88; P-values calculated by ^K-W^ do not depend on the values of the ^by^-varia > ble

^Examples:^ ^---------^

. sumby age, by(t) st(msd, range Q1 q3) dec(1 0 .) mis head . sumby age, st(msd, range Q1 q3) dec(1 0 .) . sumby age, by(t)

^Other ado files required:^ fns.ado, dummy.ado ^-------------------------^

^Author:^ ^-------^ Wim van Putten Erasmus MC - Daniel den Hoed Cancer Center Department of Statistics Rotterdam The Netherlands e-mail: w.vanputten@@erasmusmc.nl