Fw: Re: Bootstrap

[Martin Holt]



Dear Jimena,

 

I think http://www.spsstools.net/ would be a good start. If you were to translate your question/answer into English on the list, Raynald himself might help. I'll look futher later...got to take my sons to judo...unless you say you've got what you need..

 

Best Wishes,

Martin

----- Original Message -----

From: [hidden email]

To: [hidden email]

Sent: Friday, February 12, 2010 5:41 PM

Subject: Re: Re: Bootstrap

Dear MArtin

Sorry, there's no other information that I can provide... I'm sure that Hector will help us as soon as he comes back from his trip... Meanwhile, I'm probably switching to Stata because I need to finish my paper as soon as possible.

 

Thanks a lot for your compliments on my English!

 

JImena

---

From: Martin Holt <[hidden email]>
To: Jimena Maccio <[hidden email]>
Sent: Fri, February 12, 2010 2:11:28 PM
Subject: Re: Re: Bootstrap



Dear Jimena,

 

I don't know very much about SPSS/bootstrapping (that's why I posted :)), but I'm usually very good at tracking down information, though I say it myself. I'll have a go. If there is any additional information that you have that could help, I'd appreciate it if you could forward it on to me. Emails from Hector would be good. And your English is spot on......I always feel embarrassed as an Englishman that we have to ask for help from other nationalities in understanding their language, and then they're English is so often better than many English !

 

Best Wishes,

 

Martin

----- Original Message -----

From: [hidden email]

To: [hidden email]

Sent: Friday, February 12, 2010 4:24 PM

Subject: Re: Re: Bootstrap

Dear Martin,

 

I`m interested in knowing if the boostrap method is included in SPSS 17 and how to use it. I worked with Hector a couple of years ago so I consulted him about it, since he knows the program very well. He told me that there was probably a Macro I could download from the spss web site. He recommended me to visit the forum, but I'm not familiar with it so I tried finding the keyword "bootstrap" in the page but couldn`t...

 

Is there some way you can help me? Do you know how to boostrap with spss? Or where I can find this macro? Hector is leaving on a trip so he can´t help me right now...

 

Thank you VERY MUCH!!!

 

JImena

 

 

 

 

---

From: Martin Holt <[hidden email]>
To: Hector Maletta <[hidden email]>; [hidden email]; [hidden email]
Sent: Fri, February 12, 2010 1:11:29 PM
Subject: Re: Re: Bootstrap

Dear Hector and Jimena,

 

I'm interested in any posting entitled "Bootstrap" ! But I can't speak ....Spanish ? Whilst I do understand you guys communicating most accurately in your language, I'd be grateful if you could supply an English version too, if you could. Otherwise I feel like an English Manuel, from Fawlty Towers.

 

Thanks muchíiiiisimo,

 

Martin Holt

 

---- Original Message -----

From: [hidden email]

To: [hidden email]

Sent: Friday, February 12, 2010 12:38 PM

Subject: Re: Bootstrap

Jimena,

Sé que existe pero no lo usé nunca. Creo que no es un commando standard sino un macro que deberia ser bajable del sitio spss.com, en la sección Developer. Te recomendaría mirar los archivos del foro [hidden email], que creo que está archivada en http://vm.marist.edu/htbin/wlvindex?SPSSX-L. Busca bootstrap allí. Lamentablemente no tengo tiempo ahora pues estoy terminando un trabajo y luego debo ir a dejar mi laptop en el service. Mañana estoy viajando por dos semanas.

Normalmente no se usan 200 replicaciones, lo más frecuente son 10,000…….

 

HM

 

---

From: Jimena Maccio [mailto:[hidden email]]
Sent: 12 February 2010 09:25
To: [hidden email]
Subject: Bootstrap

 

Estimado Héctor,

 

Como te comenté el otro día, cuando hablamos por el tema de Haiti yo había estado a punto de llamarte para consultarte acerca del bootstrap con SPSS.

 

Hace unos meses escribí un paper para el Congreso de la AAEP (Asociación Argentina de Economía Política) en el que utilizaba datos de la EDSA para hacer un diagnóstico de la desigualdad. Calculé un conjunto de coeficientes (Gini, atkinson, etc) para los años 2004 a 2008. En ese momento lo presenté como versión preliminar porque me faltaba evaluar la sigificatividad de los cambios temporales en la desigualdad. AHora se me presenta la oportunidad de mandar el mismo paper a LACEA y me gustaría incluir estos cálculos para que el trabajo esté terminado.

 

El tema es que no sé cómo usar el spss para hacer bootstrap. Podría hacerlo de forma "manual", pero leí que son necesarios por lo menos 200 remuestreos con reemplazamiento para la estimación de los errores estándar. Dispongo de una copia del spss 17 y quería consultarte si vos conocés algún comando que me pueda asistir en este procedimiento (busqué pero realmente no puedo encontrar si existe o no)

 

Te agradezco muchíiiiisimo

Jimena

 

 

[Marta Garcia-Granero]

Hi Martin and Jimena:

I hope you don't mind if I step into this conversation. Since Hector
helped me a lot in the past, I'll be glad to help you, if I can.

To my knowledge, SPSS/PASW 17 doesn't have built-in boostrapping
capabilities, but SPSS/PASW 18 does.

Anyway, boostrapping is not that difficult, provided you have some good
tools for resampling your data a good number of times (as Hector
mentioned in your first message, it is better to resample a lot of
times, the higher the better), and you are able to compute the statistic
of interest (R-square, mean, ROC AUC, Gini coefficient) for every
sample, then use the list of statistics to:

1) Compute the SD of the statistic
2) Sort them and compute percentiles of interest (for instance,
P2.5&P97.5 would give a nive 95% boostrap CI)

As Martin told you, www.spsstools.net has some code for boostrapping,
take a look at the syntax section. I also have some tools (taking a
different approach, by using MATRIX language both for resampling and
computing the boostrapped statistics) for boostrapping the mean, ROC AUC
and their differences... You can find them here:

http://gjyp.nl/marta/


Perhaps you can use them as a starting point, since the resampling
module will be quite constant, and you would have to modify only the
section where the statistic of interest is computed.

HTH,
Marta García-Granero

PS: with Martin's permission, Jimena, estaré encantada de ayudarte en lo
que pueda. Un cordial saludo desde España


Martin Holt wrote:

> Dear Jimena,
>
> I think http://www.spsstools.net/ would be a good start. If you were
> to translate your question/answer into English on the list, Raynald
> himself might help. I'll look futher later...got to take my sons to
> judo...unless you say you've got what you need..
>
> Best Wishes,
> Martin
> ----- Original Message -----
> *From:* Jimena Maccio <mailto:[hidden email]>
> *To:* Martin Holt <mailto:[hidden email]>
> *Sent:* Friday, February 12, 2010 5:41 PM
> *Subject:* Re: Re: Bootstrap
>
> Dear MArtin
> Sorry, there's no other information that I can provide... I'm sure
> that Hector will help us as soon as he comes back from his trip...
> Meanwhile, I'm probably switching to Stata because I need to finish my
> paper as soon as possible.
>
> Thanks a lot for your compliments on my English!
>
> JImena
>
> ------------------------------------------------------------------------
> *From:* Martin Holt <[hidden email]
> <mailto:[hidden email]>>
> *To:* Jimena Maccio <[hidden email]>
> *Sent:* Fri, February 12, 2010 2:11:28 PM
> *Subject:* Re: Re: Bootstrap
>
> 
> Dear Jimena,
>
> I don't know very much about SPSS/bootstrapping (that's why I posted
> :)), but I'm usually very good at tracking down information, though I
> say it myself. I'll have a go. If there is any additional information
> that you have that could help, I'd appreciate it if you could forward
> it on to me. Emails from Hector would be good. And your English is
> spot on......I always feel embarrassed as an Englishman that we have
> to ask for help from other nationalities in understanding their
> language, and then they're English is so often better than many English !
>
> Best Wishes,
>
> Martin
>
>     ----- Original Message -----
>     *From:* Jimena Maccio <mailto:[hidden email]>
>     *To:* Martin Holt <mailto:[hidden email]>
>     *Sent:* Friday, February 12, 2010 4:24 PM
>     *Subject:* Re: Re: Bootstrap
>
>     Dear Martin,
>
>     I`m interested in knowing if the boostrap method is included in
>     SPSS 17 and how to use it. I worked with Hector a couple of years
>     ago so I consulted him about it, since he knows the program very
>     well. He told me that there was probably a Macro I could download
>     from the spss web site. He recommended me to visit the forum, but
>     I'm not familiar with it so I tried finding the keyword
>     "bootstrap" in the page but couldn`t...
>
>     Is there some way you can help me? Do you know how to boostrap
>     with spss? Or where I can find this macro? Hector is leaving on a
>     trip so he can´t help me right now...
>
>     Thank you VERY MUCH!!!
>
>     JImena
>
>
>
>
>
>
>     ------------------------------------------------------------------------
>     *
>     *
>
>

--
For miscellaneous SPSS related statistical stuff, visit:
http://gjyp.nl/marta/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Rafael raudes]

Download SPSS 18 (now it's PASW) it is capable of doing bootstrapping.


Hope it helps.

Rafael Raudes

PS: I speak spanish if you want to talk to me

2010/2/15 García-Granero <[hidden email]>

Hi Martin and Jimena:

I hope you don't mind if I step into this conversation. Since Hector
helped me a lot in the past, I'll be glad to help you, if I can.

To my knowledge, SPSS/PASW 17 doesn't have built-in boostrapping
capabilities, but SPSS/PASW 18 does.

Anyway, boostrapping is not that difficult, provided you have some good
tools for resampling your data a good number of times (as Hector
mentioned in your first message, it is better to resample a lot of
times, the higher the better), and you are able to compute the statistic
of interest (R-square, mean, ROC AUC, Gini coefficient) for every
sample, then use the list of statistics to:

1) Compute the SD of the statistic
2) Sort them and compute percentiles of interest (for instance,
P2.5&P97.5 would give a nive 95% boostrap CI)

As Martin told you, www.spsstools.net has some code for boostrapping,
take a look at the syntax section. I also have some tools (taking a
different approach, by using MATRIX language both for resampling and
computing the boostrapped statistics) for boostrapping the mean, ROC AUC
and their differences... You can find them here:

http://gjyp.nl/marta/


Perhaps you can use them as a starting point, since the resampling
module will be quite constant, and you would have to modify only the
section where the statistic of interest is computed.

HTH,
Marta García-Granero

PS: with Martin's permission, Jimena, estaré encantada de ayudarte en lo
que pueda. Un cordial saludo desde España


Martin Holt wrote:

Dear Jimena,

I think http://www.spsstools.net/ would be a good start. If you were
to translate your question/answer into English on the list, Raynald
himself might help. I'll look futher later...got to take my sons to
judo...unless you say you've got what you need..

Best Wishes,
Martin
----- Original Message -----

*From:* Jimena Maccio <mailto:[hidden email]>
*To:* Martin Holt <mailto:[hidden email]>
*Sent:* Friday, February 12, 2010 5:41 PM
*Subject:* Re: Re: Bootstrap

Dear MArtin
Sorry, there's no other information that I can provide... I'm sure
that Hector will help us as soon as he comes back from his trip...
Meanwhile, I'm probably switching to Stata because I need to finish my
paper as soon as possible.

Thanks a lot for your compliments on my English!

JImena

------------------------------------------------------------------------
*From:* Martin Holt <[hidden email]

<mailto:[hidden email]>>

*To:* Jimena Maccio <[hidden email]>
*Sent:* Fri, February 12, 2010 2:11:28 PM
*Subject:* Re: Re: Bootstrap


Dear Jimena,

I don't know very much about SPSS/bootstrapping (that's why I posted
:)), but I'm usually very good at tracking down information, though I
say it myself. I'll have a go. If there is any additional information
that you have that could help, I'd appreciate it if you could forward
it on to me. Emails from Hector would be good. And your English is
spot on......I always feel embarrassed as an Englishman that we have
to ask for help from other nationalities in understanding their
language, and then they're English is so often better than many English !

Best Wishes,

Martin

   ----- Original Message -----

   *From:* Jimena Maccio <mailto:[hidden email]>
   *To:* Martin Holt <mailto:[hidden email]>
   *Sent:* Friday, February 12, 2010 4:24 PM
   *Subject:* Re: Re: Bootstrap

   Dear Martin,

   I`m interested in knowing if the boostrap method is included in
   SPSS 17 and how to use it. I worked with Hector a couple of years
   ago so I consulted him about it, since he knows the program very
   well. He told me that there was probably a Macro I could download
   from the spss web site. He recommended me to visit the forum, but
   I'm not familiar with it so I tried finding the keyword
   "bootstrap" in the page but couldn`t...

   Is there some way you can help me? Do you know how to boostrap
   with spss? Or where I can find this macro? Hector is leaving on a
   trip so he can´t help me right now...

   Thank you VERY MUCH!!!

   JImena

   ------------------------------------------------------------------------
   *
   *

--
For miscellaneous SPSS related statistical stuff, visit:
http://gjyp.nl/marta/

=====================

To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD