Extension to cs command for exposures with more than two categories -------------------------------------------------------------------
^csjl^ case_var exp_var [weight] [^if^ exp] [^in^ range] [ ^, rr rd at type(^[ ^rr rd at^ ]^) r^ef^() c^ase^() mis^sing ^by(^varlist^ > )^ ^l^evel^()^ ^subt^ab ^subp^ct ^tab^le ^tabp^ct ^chi ex^act ^dec() to^tal > ] ^fweights^ allowed; see ^help^ @weights@.
Description -----------
^csjl^ is an extension of the ^cs^ command (for cohort studies where all subjec > ts are followed for the same length of time) that produces risk estimates and count tables for exposure variables with more than two levels. The analysis can additionally be reported by substratum defined by a ^by()^ variable. Risk estimates can be risk ratio (^rr^), risk difference (^rd^) or attributable risk > (^at^). All selected estimates shown on the same line. Testing of homogenity i > n the individual 2x2 tables can also be performed using the ^chi^ or ^exact^ opti > ons.
Options -------
^rr^ requests risk ratios (default option if no risk estimate options are specified) ^rd^ requests risk differences ^at^ requests attributable risk fraction among exposed (prevented) ^type()^ an alternative way to request one or more of the risk estimates (use this if you want to specify the output order, with spaces between entries. This option overrides options ^rr^, ^rd^ and ^at^)
^chi^ requests a Chi-sq test for subtables ^exact^ requests Fisher's exact test (two-sided) for subtables (specification of ^exact^ overrides the ^chi^ option)
^case()^ indicates group for cases in case_var (1 or 2, default: 2) ^ref()^ exposure group to be used as reference group (1...k, default: 1) ^missing^ includes missing values as a group in exposure variable ^level()^ specifies level for confidence interval for risk estimates (default: $S_level) ^dec()^ specifies number of decimal places in output values (default: 2) ^tot^ requests display of total N and # of cases in the output
^by()^ specifies variables that define substratum (multiple variables are processed sequentially and are not crossed)
^table^ requests an overall count table ^tabpct^ requests an overall count table with row percentages ^subtab^ requests count tables for the overall data and for substrata defined by the ^by()^ variable(s) ^subpct^ requests count tables with row percentages for the overall data and for substrata defined by the ^by()^ variable(s) (the table options also report an overall Chi-sq test if ^chi^ is specified) Examples --------
A minimal command yields output of the following form:
. ^csjl icd kmgrp^
Warning: no risk estimates specified, rr assumed
icd Cancer kmgrp Distance
Outcome: icd Risk Ratio ---------------------------------------------------- kmgrp N cases rr [ 95% CI ] ---------------------------------------------------- 0-25 km 1208 203 ref. group 26-62 km 1434 237 0.98 0.83 1.17 63-120 km 1942 588 1.80 1.56 2.08 120+ km 2418 986 2.43 2.12 2.78 ----------------------------------------------------
When multiple estimates are requested, they are placed side by side. The user might wish to widen the display and log line sizes:
. ^set log linesize 100^ . ^set display linesize 100^ . ^csjl icd kmgrp, type(rd rr)^
icd Cancer kmgrp Distance
Outcome: icd Risk Difference Risk Ratio ---------------------------------------------------------------------- kmgrp N cases rd [ 95% CI ] rr [ 95% CI ] ---------------------------------------------------------------------- 0-25 km 1208 203 ref. group 26-62 km 1434 237 -0.00 -0.03 0.03 0.98 0.83 1.17 63-120 km 1942 588 0.13 0.11 0.16 1.80 1.56 2.08 120+ km 2418 986 0.24 0.21 0.27 2.43 2.12 2.78 ----------------------------------------------------------------------
Options ^ref()^, ^dec()^, ^exact^, ^total^ and ^missing^ modify the output: ^ref(3)^ specifies that the 3rd level of the exposure variable is the reference exposure group; ^dec(3)^ requests 3 digits in the output; ^exact^ requests that Fisher's exact test be displayed; ^total^ causes the total counts to be displayed at the bottom of the report; and ^missing^ requests that the group of missing data in the exposure variable be tabulated as well.
. ^csjl icd kmgrp, rr ref(3) dec(3) exact total missing^
icd Cancer kmgrp Distance
Outcome: icd Risk Ratio -------------------------------------------------------------- kmgrp N cases rr [ 95% CI ] exact p -------------------------------------------------------------- 0-25 km 1208 203 0.555 0.481 0.640 0.000 26-62 km 1434 237 0.546 0.477 0.624 0.000 63-120 km 1942 588 ref. group 120+ km 2418 986 1.347 1.240 1.463 0.000 . 21 12 1.887 1.295 2.750 0.015 totals 7023 2026 --------------------------------------------------------------
Option ^case()^ specifies the level of the case_var to be used as the case group and ^chi^ requests a chi-square statistic:
. ^csjl icd kmgrp, rr ref(3) missing total dec(3) case(1) chi^
icd Cancer kmgrp Distance
Outcome: icd Risk Ratio ------------------------------------------------------------------ kmgrp N cases rr [ 95% CI ] Chi2 p ------------------------------------------------------------------ 0-25 km 1208 1005 1.193 1.148 1.240 71.9 0.000 26-62 km 1434 1197 1.197 1.153 1.243 84.5 0.000 63-120 km 1942 1354 ref. group 120+ km 2418 1432 0.849 0.813 0.888 51.5 0.000 . 21 9 0.615 0.375 1.008 7.1 0.008 totals 7023 4997 ------------------------------------------------------------------
One or more ^by()^ variables can be specified, as can a frequency weight. An unstratified table is presented first, followed by tables for each substratum of a ^by()^ variable. Note that stratification is ONLY on one ^by()^ variable at a time. . ^csjl icd kmgrp [fw=wt], rr by(sex)^
icd Cancer kmgrp Distance
Outcome: icd Risk Ratio ---------------------------------------------------- kmgrp N cases rr [ 95% CI ] ---------------------------------------------------- unstratified 0-25 km 1208 203 ref. group 26-62 km 1434 237 0.98 0.83 1.17 63-120 km 1942 588 1.80 1.56 2.08 120+ km 2418 986 2.43 2.12 2.78
by sex (Gender)
sex == unknown 0-25 km 102 90 ref. group 26-62 km 157 123 0.89 0.80 0.99 63-120 km 512 456 1.01 0.93 1.09 120+ km 867 789 1.03 0.96 1.11
sex == male 0-25 km 675 68 ref. group 26-62 km 911 93 1.01 0.75 1.36 63-120 km 954 109 1.13 0.85 1.51 120+ km 1080 153 1.41 1.07 1.84
sex == female 0-25 km 431 45 ref. group 26-62 km 366 21 0.55 0.33 0.91 63-120 km 476 23 0.46 0.28 0.75 120+ km 471 44 0.89 0.60 1.33 ----------------------------------------------------
Specification of a table option (^table^, ^tabpct^, ^subtab^, and ^subpct^) invokes Stata's ^tabulate^ command to add a count table to the output. Option ^chi^ also affects the tabulation output:
. ^csjl icd kmgrp [fw=wt], rr table chi exact^
icd Cancer kmgrp Distance
table is: | Cancer Distance | No Yes | Total -----------+----------------------+---------- 0-25 km | 1005 203 | 1208 26-62 km | 1197 237 | 1434 63-120 km | 1354 588 | 1942 120+ km | 1432 986 | 2418 -----------+----------------------+---------- Total | 4988 2014 | 7002
Pearson chi2(3) = 361.6058 Pr = 0.000
Outcome: icd Risk Ratio -------------------------------------------------------------- kmgrp N cases rr [ 95% CI ] exact p -------------------------------------------------------------- 0-25 km 1208 203 ref. group 26-62 km 1434 237 0.98 0.83 1.17 0.875 63-120 km 1942 588 1.80 1.56 2.08 0.000 120+ km 2418 986 2.43 2.12 2.78 0.000 --------------------------------------------------------------
Authors ------- Jens M. Lauritsen, jm.lauritsen@@dadlnet.dk County of Funen, Odense, Denmark
Thomas J. Steichen, steicht@@rjrt.com Winston-Salem, NC, USA
Also see --------- On-line: help for @epitab@, @cs@, @set display linesize@, @csi@, @level@