..- help for ^cid^ (statalist Mar. 20, 1998) ..- Confidence intervals for differences - ------------------------------------ ^cid^ varname [^if^ exp] [^in^ range] [^, m^edian ^l^evel^(^#^)^ ] ^cid^ varname1 varname2 [^if^ exp] [^in^ range] [^, m^edian ^unp^aired ^une^qual ^l^evel^(^#^)^ ] ^cid^ varname [^if^ exp] [^in^ range]^, by(^groupvar^)^ [ ^m^edian ^unp^aired ^une^qual ^l^evel^(^#^)^ ] ^cid^ calculates confidence intervals (CIs) for means or differences. In the first form, ^cid^ calculates a CI for the mean of varname. In the second form, ^cid^ produces a CI for mean difference between varname1 and varname2. Data are assumed to be paired, but ^unpaired^ changes this assumption. In the third form, ^cid^ gives a CI for the mean difference across the two groups defined by groupvar. Data are assumed paired, but ^unpaired^ reverses this. Options - ------- ^median^ calculates CI's as above, except for medians rather than for means. CI's are determined by rank-based methods as described in Gardner and Altman's (1989) book "Statistics with Confidence", pp. 74-79. A normal approximation with a continuity correction of 0.5 is used, rather than tables of exact values. The accuracy should be adequate when both groups have at least 5 observations. Note that the CI for a single median is obtained using a method different from that used by the @centile@ command, so will not be identical. WARNING: calculation of CIs for unpaired medians requires storage space for n1 * n2 observations, where n1 and n2 are the numbers in each group. If you have insufficient memory, you will get an error message and a ^r(2002);^ return code. ^unpaired^ indicates that the data are to be treated as unpaired. ^unequal^ indicates that the unpaired data are not to be assumed to have equal variances. ^level(^#^)^ defines the percent coverage for the confidence interval. Examples - -------- . ^cid bp2 bp1^ . ^cid bp2 bp1, median^ . ^cid bp2 bp1, median unpaired^ . ^cid bp, by(treatmnt)^ . ^cid bp, by(treatmnt) median^ . ^cid bp, by(treatmnt) unpaired^ . ^cid bp, by(treatmnt) unpaired unequal^ Stored - ------ ^\$S_1^ number of observations ^\$S_2^ degrees of freedom (not used if ^median^ is specified) ^\$S_3^ mean or median difference, group 1 - group 2 ^\$S_4^ SEM, or K if ^median^ specified (see Gardner and Altman) ^\$S_5^ lower confidence limit for difference ^\$S_6^ upper confidence limit for difference Also see - -------- Manual: [R] @ci@, @centile@, @ttest@, @ranksum@, @signrank@.