st: SSC new stuff

[Adrian Mander <Adrian.Mander@mrc-hnr.cam.ac.uk>]

Hi Kit,

I have written a new samplesize calculation for matched case/control studies.
called sampsi_mcc.

I have also had to embed this in my samplesize.ado file... 
Essentially I am trying to
write the whole of PASS2005 or encourage others to have a one stop 
shop for sample size and
power.

Anyway here are the files.

cheers
Ade
{smcl}
{hline}
help for {hi:samplesize}
{hline}
{title:Executes sample size/power calculations multiple times and produces graphical results}

{p 8 22}
{cmdab:samplesize}
[,
{opt null(numlist)}
{opt alt(numlist)}
{opt n1(numlist)}
{opt n2(numlist)}
{opt sd1(numlist)}
{opt sd2(numlist)}
{opt rho(numlist)}
{opt a:lpha(numlist)}
{opt p:ower(numlist)}
{opt s:olve(string)}
{opt r:atio(numlist)}
{opt xvar(string)}
{opt onesam:ple}
{opt onesided}
{opt me:thod(string)}
{opt nocont:inuity}
{opt pre(numlist)}
{opt post(numlist)}
{opt r0(numlist)}
{opt r1(numlist)}
{opt r01(numlist)}
{opt sy(numlist)}
{opt sx(numlist)}
{opt yxcorr(numlist)}
{opt var:method(string)}
{opt command(string)}
{opt m(numlist)}
{opt phi(numlist)}
{opt p0(numlist)}
{help twoway_options }
]

{p}

{title:Description}

{p 0 0}
Most of STATA's sample size calculation programs do not allow {hi:numlists} for the options.
{hi: samplesize} is designed to allow {hi:numlists} to do multiple calculations using various sample size commands.  
The resulting sample sizes or power calculations are then drawn using a {hi:twoway} graph.

{p 0 0}
At present the following commands are supported (more will be introduced):

Help File                      Examples
{help sampsi}                {help samplesize##ttest:Two-sample t-test}
{help sampsi_reg}            {help samplesize##linreg:Linear regression }
{help sampsi_mcc}            {help samplesize##mcc:Matched Case-Control }
{* help mvsampsi              help samplesize##mv:Multivariate Regression }

Please email me if you want the introduction of other sample size commands.
 
{title:Updating this command}

{p 0 0}
To obtain the latest version click the following to uninstall the old version
{p_end}
{stata ssc uninstall samplesize}
And click here to install the new version
{stata ssc install samplesize}

{title:Options}

{p 0 0}
{opt null(numlist)} specifies the "null value", #1 in {hi:sampsi}.

{p 0 0}
{opt alt(numlist)} specifies the "alternative value", #2 in {hi:sampsi}.

{p 0 0}
{opt n1(numlist)} size of sample 1. For {hi:sampsi_mcc} this is the number of cases.

{p 0 0}
{opt n2(numlist)} size of sample 2.

{p 0 0}
{opt sd1(numlist)} standard deviation of sample 1.

{p 0 0}
{opt sd2(numlist)} standard deviation of sample 2.

{p 0 0}
{opt a:lpha(numlist)} significance level of test; default is {hi:a(0.05)}.

{p 0 0}
{opt p:ower(numlist)} power of test; default is {hi:p(0.9)}.

{p 0 0}
{opt s:olve(string)} specifies whether to solve for the sample size or power; default is {hi:s(n)} solves for n and
the only other choice is {hi:s(power)} solves for power.

{p 0 0}
{opt r:atio(numlist)} ratio of sample sizes; default is {hi:r(1)}.

{p 0 0}
{opt xvar(string)} specifies the variable to be used as the x-variable in the resulting plots. 
The default is the variable with the most values, this will work well for the majority of calculations.

{p 0 0}
{opt onesam:ple} one-sample test; default is two-sample.

{p 0 0}
{opt onesided} one-sided test; default is two-sided.

{p 0 0}
{opt m:ethod(string)} analysis method is {hi:post}, {hi:change}, {hi:ancova}; default is {hi:m(all)} 
although only {hi:ancova} will be plotted.

{p 0 0}
{opt nocont:inuity} do not use continuity correction for two-sample test on proportions.

{p 0 0}
{opt pre(numlist)} number of baseline measurements; default is {hi:pre(0)}.

{p 0 0}
{opt post(numlist)} number of follow-up measurements; default is {hi:post(1)}.

{p 0 0}
{opt r0(numlist)} correlation between baseline measurements; default is {hi:r0(0)}.

{p 0 0}
{opt r1(numlist)} correlation between follow-up measurements; default is {hi:r1(0)}.

{p 0 0}
{opt r01(numlist)} correlation between baseline and follow-up measurements; default is {hi:r01(0)}.

{p 0 0}
[{help sampsi_reg} option] {opt sy(numlist)} the standard deviation of the Y's.

{p 0 0}
[{help sampsi_reg} option] {opt sx(numlist)} the standard deviation of the X's.

{p 0 0}
[{help sampsi_reg} option] {opt yxcorr(numlist)} the correlation between Y's and X's.

{p 0 0}
[{help sampsi_reg} option] {opt var:method(string)} specifies the method for calculating the residual standard deviation. 
{opt varmethod(r)} uses the Y-X correlation and  {opt varmethod(sdy)} uses the standard deviation of the Y's,
the default uses a direct estimate of the residual sd {opt sd1(#)}. 

{p 0 0}
[{help sampsi_mcc} option] {opt m(numlist)} specifies the number of matched controls per case; default is {hi:m(1)}.

{p 0 0}
[{help sampsi_mcc} option] {opt phi(numlist)} specifies the correlation of exposure between pairs of subjects in the 
case-control matched set; default is {hi:phi(0.2)}.

{p 0 0}
[{help sampsi_mcc} option] {opt p0(numlist)} specifies the probability of exposure in the controls; default is {hi:p0(0.5)}.

{p 0 0}
{opt command(string)} specifies the sample size command, the default is {hi:sampsi}.


{title:Examples}

{p 0 0}
The full interactive version runs from a dialog box
{stata db samplesize} (to be distributed in the near future)

{marker ttest}
{p 0 2}
Two-sample comparison of mean1 to mean2.  Compute sample sizes with n2/n1 = 2:
{p_end}
{p 2 2}
{stata samplesize, null(132.86) alt(127.44) p(0.8) r(2(2)10) sd1(15.34) sd2(18.23)}
{break}
Compute power with n1 = n2, sd1 = sd2, and alpha = 0.01 one-sided:
{break}
{stata samplesize, null(5.6) alt(6.1) n1(100) sd1(1.5) a(0.01(0.01)0.05) onesided }

{p 0 2}
One-sample comparison of mean to hypothesized value = 180.  Compute sample size:
{p_end}
{p 2 2}
{stata samplesize, null(180) alt(211) sd(46(1)60) onesam }
{break}
One-sample comparison of mean to hypothesized value = 0.  Compute power:
{break}
{stata samplesize, null(0) alt(-2.5) sd(4(0.2)5) n(25(10)55) onesam }

{p 0 2}
Two-sample comparison of proportions.  Compute sample size with n1 = n2 (i.e., ratio = 1, the
default) and power = 0.9 (the default):
{p_end}
{p 2 2}
{stata samplesize, null(0.25) alt(0.4(0.01)0.6)}
{break}
Compute power with n1 = 500 and ratio = n2/n1 = 0.5:
{break}
{stata samplesize, null(0.25) alt(0.4) n1(300) r(0.5(0.1)0.9) }

{p 0 2}
One-sample comparison of proportion to hypothesized value = 0.5:
{p_end}
{p 2 2}
{stata samplesize, null(0.5) alt(0.75) power(0.8(0.01)0.9) onesample }
{break}
Compute power:
{break}
{stata samplesize, null(0.5) alt(0.6) n1(200(10)400) onesam s(power)}

{p 0 2}
Repeated Measures
{p_end}
{p 2 2}
{stata samplesize, null(498) alt(483(0.2)487) sd1(20.2) sd2(19.5) method(change) pre(1) post(3) r1(.1(.1).9) solve(n) }
{break}
Compute power:
{break}
{stata samplesize, null(498) alt(485) sd1(20.2) sd2(19.5) method(change) pre(1) post(1(1)10) r1(.7) n1(15) n2(15) solve(power)}

{marker linreg}
{p 0 2}
Linear Regression
{p_end}
{p 2 2}
{stata samplesize, null(0) alt(0.2(0.1)0.8) solve(n) command(sampsi_reg) }
{break}
Compute power:
{break}
{stata samplesize, null(0) alt(0.6(0.1)1.6) sx(0.5(0.2)1.5) solve(power) command(sampsi_reg)}

{marker mcc}
{p 0 2}
Matched Case-control Study.
{p_end}
{p 2 2}
{stata samplesize, alt(1.2(0.1)1.8) m(1(1)5) solve(n) command(sampsi_mcc) }
{break}
Compute power:
{break}
{stata samplesize, alt(1.2(0.1)3) phi(0.2(0.2)0.8) n1(100) solve(power) command(sampsi_mcc)}

{title:Author}

{p}
Adrian Mander, MRC Human Nutrition Research, Cambridge, UK.

Email {browse "mailto:adrian.mander@mrc-hnr.cam.ac.uk":adrian.mander@mrc-hnr.cam.ac.uk}

{title:See Also}
Related commands:

{p 2 2}
{help sampsi}, 
{help sampsi_reg} (if installed),
{help sampsi_mcc} (if installed),
{help sampclus} (if installed), 
{help xsampsi} (if installed), 
{help artmenu} (if installed),
{help mvsampsi} (if installed),
{help studysi} (if installed),
{help sskapp} (if installed),
{help ssizebi} (if installed),
{help optfixn} (if installed),
{help calcssi} (if installed),
{help ggipower} (if installed),
{help sampncti} (if installed).

Attachment: samplesize.ado
Description: Binary data

{smcl}
{hline}
help for {hi:sampsi_mcc}
{hline}

{title:Calculates Sample Size or Power for Matched Case-Control Studies}

{p 8 22}
{cmdab:sampsi_mcc}
[,
{opt p0(#)}
{opt alt(#)}
{opt n1(#)}
{opt m(#)}
{opt phi(#)}
{opt a:lpha(#)}
{opt power(#)}
{opt s:olve(string)}
}
]

{p}

{title:Description}

{p 0 0}
{hi: sampsi_mcc} calculates the power and sample size for a matched case control study. The theory behind this
command is described in Dupont (1988) Power Calculations for Matched Case-Control Studies, Biometrics.

{p 0 0}
The calculations require the usual alpha and beta values, a possible alternative odds ratio (the null is 1), phi the
correlation of exposure between pairs in the case-control set (the default is 0.2) and the probability of exposure in
the controls.

{p 0 0}
This command can be combined with {hi:samplesize} in order to look at multiple calculations and to plot the results.

{title:Updating this command}

{p 0 0}
To obtain the latest version click the following to uninstall the old version
{p_end}
{stata ssc uninstall sampsi_mcc}
And click here to install the new version
{stata ssc install sampsi_mcc}

{title:Options}

{p 0 0}
{opt p0(#)} specifies the exposure probability of controls; default is {hi:p0(0.5)}.

{p 0 0}
{opt alt(#)} specifies the "alternative OR".

{p 0 0}
{opt n1(#)} specifies the number of cases; default is {hi:n1(100)}.

{p 0 0}
{opt m(#)} specifies the number of matched controls per case; default is {hi:m(1)}.

{p 0 0}
{opt a:lpha(#)} significance level of test; default is {hi:a(0.05)}.

{p 0 0}
{opt power(#)} power of test; default is {hi:p(0.9)}.

{p 0 0}
{opt s:olve(string)} specifies whether to solve for the sample size or power; default is {hi:s(n)} solves for n and
the only other choice is {hi:s(power)} solves for power.

{p 0 0}
{opt phi(#)} specifies the correlation of exposure between pairs in the case-control set. The paper recommends the
default value {hi:0.2} when it is unknown.

{title:Examples}

{p 0 0}
Calculate power for a two-sided test:
{p_end}

{p 2 2}
{stata sampsi_mcc, p0(0.5) alt(1.5) n(100) s(power)}

{p 0 0}
Compute sample size:

{p 2 2}
{stata sampsi_mcc, p0(0.5) alt(1.5) s(n)}

{p 0 0}
Re-calculate the above values for 5 matched controls rather than 1:
{p_end}

{p 2 2}
{stata sampsi_mcc, p0(0.5) alt(1.5) n(100) s(power) m(5)}

{p 0 0}
Compute sample size:

{p 2 2}
{stata sampsi_mcc, p0(0.5) alt(1.5) s(n) m(5)}

{title:Author}

{p}
Adrian Mander, MRC Human Nutrition Research, Cambridge, UK.

Email {browse "mailto:adrian.mander@mrc-hnr.cam.ac.uk":adrian.mander@mrc-hnr.cam.ac.uk}

{title:See Also}
Related commands:

{p 2 2}
{help sampsi}, {help samplesize} (if installed), {help sampclus} (if installed), {help xsampsi} (if installed), {help artmenu} (if installed)

Attachment: sampsi_mcc.ado
Description: Binary data