Dear Christian,
sorry for my previous mistyping your name.
Kind Regards,
Carlo
-----Messaggio originale-----
Da: [email protected]
[mailto:[email protected]] Per conto di Carlo Lazzaro
Inviato: lunedì 6 luglio 2009 9.32
A: [email protected]
Cc: 'Christian Bustamante'
Oggetto: R: st: R:R:st Mata question
Dear Christina,
thanks a lot for your kind reply.
My temptative answer was admittedly not in Mata language, since I am not a
Mata expert: my intent was to give a little contribution about different
paths for solving your problem.
Besides:
- I might have been unclear in my previous pots but the .do file generates
mean and sds (and not observations) of 20 different samples (10 observations
each) drawn from two population, supposed to have the same variance;
- more important: the seventh line of the .do file from the bottom
qui g w`i'=ttail(18,abs(t`i')) should be
qui g w`i'=2*ttail(18,abs(t`i'))
As an acid test for testing the results obtained with the .do file is to
compare them to those obtained via - ttesti - command, invoked from within
Stata 9.2/SE.
Considering mean_alfa[1](= .27396816), sd_alfa[1] ](= .02923826),
mean_beta[1] ](= .2769154) and sd_beta[1] ](= .01790745), I get the
following student_t[1]= -.271828263998; p_value[1]= .788847386837.
Switching to -ttesti-:
ttesti 10 .27396816 . 02923826 10 .2769154 .01790745
Two-sample t test with equal variances
----------------------------------------------------------------------------
--
| Obs Mean Std. Err. Std. Dev. [95% Conf.
Interval]
---------+------------------------------------------------------------------
--
x | 10 .2739682 .0092459 .0292383 .2530524
.294884
y | 10 .2769154 .0056628 .0179074 .2641052
.2897256
---------+------------------------------------------------------------------
--
combined | 20 .2754418 .0052874 .0236459 .2643752
.2865084
---------+------------------------------------------------------------------
--
diff | -.0029472 .0108423 -.025726
.0198316
----------------------------------------------------------------------------
--
diff = mean(x) - mean(y) t =
-0.2718
Ho: diff = 0 degrees of freedom =
18
Ha: diff < 0 Ha: diff != 0 Ha: diff > 0
Pr(T < t) = 0.3944 Pr(|T| > |t|) = 0.7888 Pr(T > t) =
0.6056
. return list
scalars:
r(sd) = .0236458571402653
r(sd_2) = .01790745
r(sd_1) = .02923826
r(se) = .0108422904099185
r(p_u) = .6055762826325284
r(p_l) = .3944237173674716
r(p) = .7888474347349431
r(t) = -.2718281736213082
r(df_t) = 18
r(mu_2) = .2769154
r(N_2) = 10
r(mu_1) = .27396816
r(N_1) = 10
All the Best for your research project, hoping that some Mata guru can
provide you with more helpful hints.
Kind Regards,
Carlo
-----Messaggio originale-----
Da: [email protected]
[mailto:[email protected]] Per conto di Christian
Bustamante
Inviato: lunedì 6 luglio 2009 1.53
A: [email protected]
Oggetto: Re: st: R:R:st Mata question
Hi Carlo,
Thanks for your help, but I apparently it doesn't works for what I
need because I have vector with means a sds, not "observarions", thing
that is not trivial. I've continued my tryings whit this:
forvalues i=1/52 {
scalar m1=dy04[1,`i']
scalar s1=S04[1,`i']
scalar m2=dy07[1,`i']
scalar s2=S07[1,`i']
ttesti N04 m1 s2 N07 m2 s2
mat tvalues[1,`i']=r(t)
mat pvalues[1,`i']=r(p)
}
But stata propmt says: "'N04' found where integer expected" and "'m1'
found where number expected", so i ask: there is a way to convert an
scalar in a "integer" or in a "number"?
Bests regards
On Sun, Jul 5, 2009 at 5:26 AM, Carlo
Lazzaro<[email protected]> wrote:
> Dear Christian,
> the following is a a fine-tuning of my yesterday's temptative answer to
your
> query, again not in Mata language.
> Let's assume we have 20 + 20 sample means and sds (sample size: 10
> observations each) drawn from 2 different populations (I mean, 20 from
> population1 and 20 from population 2). Sample means and sds drawn from
> population 1 are now labelled mean_alfa and sd_alfa. Sample means and sds
> drawn from population 2 are now labelled mean_beta and sd_beta. Ttests are
> performed under the equal variance assumption; hence, I have plugged in
> pooled_variance formula (please, see v`i' line of the beneath reported
Stata
> 9.2/Se .do file), as from Pagano M, Gavreau K. Principles of
Biostatistics.
> 2nd edition. Brooks/Cole, 2000 (p 206 of the Italian edition).
> Example now includes retrieving and storing two-sided p-value.
>
> set obs 20
> g mean_alfa=2*uniform()
> g mean_beta=1*uniform()
> g sd_alfa=.03*uniform()
> g sd_beta=.02*uniform()
> g double var_pooled=.
> g double student_t=.
> g double p_value=.
> qui forval i = 1/20 {
> qui g v`i'=((10-1)*sd_alfa[`i']^2+(10-1)*sd_beta[`i']^2)/(10+10-2)
> sum v`i', meanonly
> replace var_pooled=r(mean) in `i'
> qui g t`i' = (mean_alfa[`i']-mean_beta[`i'])/(v`i'*(1/10+1/10))^.5
> sum t`i', meanonly
> replace student_t=r(mean) in `i'
> qui g w`i'=ttail(18,abs(t`i'))
> sum w`i', meanonly
> replace p_value=r(mean) in `i'
> drop v`i'
> drop t`i'
> drop w`i'
> }
>
> Again, not that elegant, but I do hope this help.
>
> Kind Regards,
> Carlo
> -----Messaggio originale-----
> Da: Carlo Lazzaro [mailto:[email protected]]
> Inviato: sabato 4 luglio 2009 10.46
> A: '[email protected]'
> Cc: 'Christian Bustamante'
> Oggetto: R: Mata question
>
> Dear Christian,
> the following is a temptative answer to your query, not in Mata language.
> Let's assume we have 20 + 20 sample means and sds (sample size: 10
> observations each) drawn from 2 different populations (I mean, 20 from
> population1 and 20 from population 2). Sample means and sds drawn from
> population 1 are labelled a and c. Sample means and sds drawn from
> population 2 are labelled b and d. Ttests are performed under the equal
> variance assumption. Example stops at Student t retrieving and storing for
> each ttest performed (two-sided p-value was not included in the Stata .do
> file).
>
> set obs 20
> g a==2*uniform()
> g b=1*uniform()
> g c=.03*uniform()
> g d =.03*uniform()
> g double student_t=.
> qui forval i = 1/20 {
> g t`i' = (a[`i']-b[`i'])/(c[`i']^2/10+d[`i']^2/10)^.5
> sum t`i', meanonly
> replace student_t=r(mean) in `i'
> drop t`i'
> }
>
> Not that elegant, but I do hope this help.
>
> Kind Regards,
> Carlo
> -----Messaggio originale-----
> Da: [email protected]
> [mailto:[email protected]] Per conto di Christian
> Bustamante
> Inviato: sabato 4 luglio 2009 9.15
> A: Statalist
> Oggetto: st: Mata question
>
> Hi all
>
> I'm trying to do a loot with a nested -ttesti- but I couldn't do it. I
> have two vector with different means (m1 and m2) and another two with
> each mean's standard deviation (s1 and s2). This vectors are quite
> large, so do -ttesti- element for element could be very costly. Also,
> I have each sample's size on two variables: n1 and n2. All this values
> are obtained from a logistic regresion. After run -ttesti- I want to
> save the t-statistic -r(t)- and the probability for two-sided p-value
> -r(p)-. I'm doing something like this:
>
> forvalues i=1/size(m1) { /* how can i get vector's size? */
> ttesti n1 m1[1,`i'] s1[1,`i'] n1 m1[1,`i'] s1[1,`i']
> mat tvalues[1,`i']=r(t)
> mat pvalues[1,`i']=r(p)
> }
>
> But appears a lot of errors: 1) n1 should be integer. 2) 'm1' found
> where number expected. 3) varlist not allowed.
> How can I solve this?
>
> Thanks
>
>
>
> --
> CdeB
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
--
CdeB
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
