[pymvpa] permutator and searchlight

[Nick Oosterhof]

Dear Cai Fei,

On Aug 28, 2014, at 4:52 PM, cfygj <cfygj at 126.com> wrote:

> [...] My questions in the final. [...]
> (1)My frist goal is to estimate the null-distribution of the prediction accuracy of the classifiers. Can the scripts above realise this goal?
> (2)How should I do to conduct searchlight analysis using the scripts above?
> Any advise should be appreciated! 

you've posted several times to this mailinglist now, with recently several repeats of more or less the same question.  You may have noticed that for your more recent questions there have been few, if any, replies. 

Remember some advice I gave you earlier? 

"be more specific about:

1) what you want to achieve
2) what information you have found that may help in achieving that goal
3) what you've tried
4) what output you got
5) what output you expected to get

Broad comments saying you're looking for "advice on some ways to deal with my problems" or that "some error information emerged", in combination with just pasting a script, (a) gives usually not enough information to give good advice; and (b) may convey the impression that you expect others you solve your problems without you having tried very hard yourself."

So far I don't have the impression that you followed my advice. That is your good right, of course, but it may hinder you in getting useful replies to your questions.

Let me try to be clear about how I think about this:
- there are some very clever people subscribed to this mailinglist, and these people are generally nice and helpful. You have potential access to a lot of knowledge here, but please use it wisely.
- posting the same, or almost the same, message multiple times will not make it more likely to get replies. Unless you think you have technical issues with sending messages to the mailinglist, there is no advantage of multiple posts. In fact it will make it more likely that people will start to ignore your messages: people on this mailinglist do not want to waste their time, and reading the same message multiple times for no clear reason *does* waste their time.
- do not expect that other people write your analysis scripts, unless you have a very interesting problem or dataset. What is interesting is somewhat subjective; eCoG data from Stephen Hawking would most likely qualify, but a standard fMRI analysis with training and testing on two chunks with two classes (your case) does not.
- if you want people to have a look at your scripts, do your homework first - read the documentation of the functions, try things out yourself, and attempt to find a solution first before you post a message here. 

In general your messages so far may convey an impression of both laziness and arrogance. Laziness, because you ask people how to do things for you, but give no indication that you have tried hard yourself (if you have); and if there is no response, you just post the same message again. Arrogance, because it would seem that your time is more valuable than other subscribers on this mailinglist. Even if you were lazy or arrogant, you do not want to convey such an impression in order to get useful replies here.

Before you post more messages, I strongly suggest that you read (and follow the advice in) "how to ask questions the smart way"

http://www.catb.org/~esr/faqs/smart-questions.html

Even if you are not a hacker (I don't consider myself to be one), it is worthwhile to try and act like one.

best,
Nick