Bootstrap as manual

[Akya Nus]

Hi,

I'm studying bootstrap for microarray analysis. I found this code from http://www.compgen.org/tools/microarrays

program define ttestboot, rclass
version 10.1
    syntax , x(varlist numeric max=1) type(varlist numeric max=1) [ reps(real 100) var(string )  ]
set more off

di in ye "______________________________________________________________________________"

di in ye "   Calculation of Achieved Significance Level (ASL) using the bootstrap"
di in ye "   The idea is to recenter the two samples to the combined sample mean"
di in ye "   so that the data now conform to the null hypothesis but that the"
di in ye "   variances within the samples remain unchanged"
di in ye "______________________________________________________________________________"

preserve
ttest `x',by(`type') uneq
tempname tobs omean
scalar `tobs' = r(t)
qui summarize `x', meanonly
scalar `omean' = r(mean)
qui summarize `x' if `type'==0, meanonly
qui replace `x' = `x' - r(mean) + scalar(`omean') if `type'==0
qui summarize `x' if `type'==1, meanonly
qui replace `x' = `x' - r(mean) + scalar(`omean') if `type'==1
tempfile boot

bootstrap t=r(t),nolegend nowarn notable reps(`reps') strata(`type') saving(`boot'): ttest `x',by(`type') `var'

use `boot',clear
qui generate indicator = abs(t)>=abs(scalar(`tobs'))
qui summarize indicator, meanonly
display in ye "ASLboot = " r(mean)
restore
return scalar p=r(mean)
end

I have a gene expression data which has 10 Treatment and 10 control samples with 250 genes. I converted it vectorized format. For example:

Gene	Exp	Type
Gene1	1.602014	1=treat
Gene1	1.497501	0=control
Gene2	1.675608	1
Gene2	3.727904	0

When I run this code for this table I 'm getting this error:

ttestboot Expr Type [reps(100) Gene]
#varlist not allowed

I guess I don't know and understand calling program. I described list object with this command:

list Expr Type Gene

How can I call ttestboot ? Can you suggest something or share knowledge?

[Ken Chui]

Based on line 3, the arguments you provided should be treated as options. Try something like this to start:

ttestboot, x(Expr) type(Type) reps(100)

[Akya Nus]

Thank you for ypur response. it is working now, but only one t obs. was produced. we can see all genes with type. My aim is to find expressed genes. In this code var(string ) can be genes?

[Ken Chui]

I have not worked with this command but from what I understand that "var(string)" is reserved for the option items specific to the -ttest- command. See -help ttest- and look under "option 1" in the syntax chart.

If you want to repeat that by Gene, I'd suggest:

bysort Gene: ttestboot, x(Expr) type(Type) reps(100)

[Akya Nus]

Thank you I will check.

[Akya Nus]

I tried some options from help, but I got an error.

by gene, sort : ttestboot, x(Expr) type(Type) reps(100)
ttestboot may not be combined with by

[Ken Chui]

I tried some options from help, but I got an error.

by gene, sort : ttestboot, x(Expr) type(Type) reps(100)
ttestboot may not be combined with by

I'm just going off on guesses here. First, try "if". For example:

ttestboot if Gene == "Gene1", x(Expr) type(Type) reps(100)

If it still doesn't work, then you may have to subset the data using -keep- and then apply the code.

Both of them may require some looping depending how many gene types are there, check -foreach- to learn more.

Good luck!