Statscraft 0.15.0, release v0.15.0
CSTABULATE
One-way and two-way tables of a complex sample with design-based errors and adjusted tests.
Syntax
CSTABULATE /PLAN FILE='file' /TABLES VARIABLES=vars [BY var]
[/CELLS [POPSIZE] [ROWPCT] [COLPCT] [TABLEPCT]]
[/STATISTICS [SE] [CV] [DEFF] [DEFFSQRT] [CIN[(95)]] [COUNT] [CUMULATIVE]
[EXPECTED] [RESID] [ASRESID]]
[/TEST [HOMOGENEITY] [INDEPENDENCE] [ODDSRATIO] [RELRISK] [RISKDIFF]]
[/SUBPOP TABLE=var [BY var ...] [DISPLAY=LAYERED | SEPARATE]]
[/MISSING [SCOPE=TABLE | LISTWISE] [CLASSMISSING=EXCLUDE | INCLUDE]].
Each variable gets a one-way table, or with BY a two-way table against the BY variable, with
estimated population sizes and percents for every cell and margin. The design comes from a
CSPLAN plan, as in CSDESCRIPTIVES, and so do the standard errors, df and design effects.
Percents are ratios of population sizes; their intervals are found on the logit scale.
CUMULATIVE (one-way) adds running sums. EXPECTED, RESID and ASRESID (two-way, with
TABLEPCT) show the table percent expected under independence, the difference from it and
that difference over its design-based standard error. HOMOGENEITY tests equal shares of a
one-way table, INDEPENDENCE the rows against the columns: the Pearson and likelihood ratio
chi-squares are divided by the sum of the generalized design effects and referred to F
with d and d times the design df, d being the squared sum over the sum of squares of those
effects (a second-order adjustment). ODDSRATIO, RELRISK and RISKDIFF give the 2 x 2
measures with intervals (log scale for the first two). Tests are for the whole population,
and also for each subpopulation when SUBPOP names only first-stage strata. Without CELLS:
population sizes; without STATISTICS: standard errors. Not supported yet: JOINTPROB.
Example
CSTABULATE /PLAN FILE='farms_plan.json' /TABLES VARIABLES=irrigated BY organic /CELLS POPSIZE ROWPCT /STATISTICS SE CIN DEFF /TEST INDEPENDENCE ODDSRATIO.
statscraft help CSTABULATE prints the same text in the terminal. All commands