[R] A humble request

Sun Jul 3 14:44:50 CEST 2022

This is why research students have supervisors.
A research degree is training in doing research.
When you get stuck, your supervisor (or co-supervisor) is supposed to help
you unstick yourself.
Asking random strangers can never be as good as asking someone who already
understands whatever it is you are actually trying to do.
Supervisors have knowledge and experience relevant to the research their
students are doing.  It's not always *much* more.  In fact, I once
"supervised" a student who had years of industrial experience in his topic,
when I knew just the basics of the subject.  He taught me a lot.
But one thing a supervisor ought to be able to do for you is to ask the
random strangers in a much clearer way.  He should be able to explain what
you are trying to do.
If, for example, he knows that you are trying to replicate someone else's
work as a prelude to extending it, he has the status to ask THEM what
package they recommend now.

You seem to be insisting that this one package is what you need and no
other, but you haven't told us WHY it has to be this package.
- We do not know what kind of data you have.
- We do not know why you think a wavelet analysis is appropriate.
  (I don't really understand wavelets and what they can do.  It is a big
gap in my education, but there are so many others it's near the bottom of
the list.)
- We do not yet have any reason to believe that a wavelet analysis IS a
particularly good way to deal with your data.
- We do not know what you want to learn from the data.
- We do not know whether your literature survey has turned up any other
papers doing something similar.
The metaphor I'm about to offer is a bit hyperbolic, but I hope it gets the
idea across.
Student: I want to get the HCN from the Poisons Cabinet here but the key I
have doesn't work.
Helpful random stranger: I tell you what, why not join a lock sports club
[they exist], buy the Genesis kit [real product] of lockpicking tools from
Covert Instruments [real company], and learn how to pick the lock?
REALLY helpful random stranger: What the hell do you want HCN for?

Other people have made it clear that you personally are NOT going to get
wmtsa working with reasonable effort, and nobody is paying any of us to
resurrect a dead package.  At SOME point you are going to have to describe
very clearly, as if to an intelligent 14-year-old,
- what kind of data you have
- what physical or social processes generated that data
- what kind of insight you think can be obtained from it
- what the raw data look like
- some crude "exploratory" statistics on the data
- what kinds of models other people with similar data have used
- what specific methods of analysis you intend to use/did use
- how you can tell whether the answers make sense
- what you are going to do with the answers
All of this needs to be part of the thesis.  You are going to have to write
it SOME time.  If not now, when?  If not you, who?

When you understand what you want to do well enough to explain it, you will
be able to ask much better questions, of the form
"I wanted to use wmtsa to do <X> to <Y>.
 What's a good package for that these days?
 Is there a textbook about doing <X> with R?"

As a former supervisor, I hope, oh how much I hope, that you were not
planning to just load wmtsa, turn the crank on a script you found down the
back of a couch, and believe whatever came out.  I've known Masters and
even PhD students do that.  Not mine, at least not for long.  I also hope
that you have a good supervisor and a good working relationship.  Speaking
from experience, one thing that really ANNOYS a good supervisor is a
student who stays stuck instead of asking for advice.  (I've been that
student.  I was much stupider then.)  It makes the supervisor feel
disrespected.

On Sun, 3 Jul 2022 at 18:00, Muhammad Zubair Chishti <
mzchishti using eco.qau.edu.pk> wrote:

> Dear Respected Experts and specifically Professor Richard O'Keefe,
> Thank you so much for your precious time and generous help. However, the
> problem is still there and I am just unable to resolve it due to the lack
> of expertise in R. Still, the hope is there. I believe that this platform
> can help me.
>
> Regards
> Muhammad Zubair Chishti
> School of Business,
> Zhengzhou University, Henan, China
> My Google scholar link:
> https://scholar.google.com/citations?hl=en&user=YPqNJMwAAAAJ
> My ReseachGate Profile:
> https://www.researchgate.net/profile/Muhammad-Chishti
>
>
> On Sun, Jul 3, 2022 at 9:11 AM Richard O'Keefe <raoknz using gmail.com> wrote:
>
>> A postscript to my previous comment.
>> I used to supervise PhD students.
>> Let me advise you to write this issue up as a draft section for your
>> thesis.
>> 1. Why I wanted to use the wmtsa package.
>> 2. Why I didn't.
>> 3. How I went about selecting a replacement.
>> 4. What I chose and why that's the right choice.
>> 5. How the analyses I wanted to do are done in
>>    package X and what difference it makes.
>>
>> Off the top of my head, the only reasons for struggling to use an old
>> package are to try to replicate someone else's results and/or to try to use
>> their software (built atop the dead package) with new data.  Well, if you
>> get different results, that's interesting too, and then it's time to work
>> harder to resurrect the dead package.
>>
>> Speaking of which, an easier route might be to set up a separate
>> environment running an old version of R that *can* run the old code and the
>> old code's dependencies.  In fact trying to use the same versions that the
>> work you're trying to reproduce used might make a lot of sense.
>>
>> Overall, I think selecting an alternative package
>> that *is* currently maintained is the best use of your time, but your
>> supervisor should be able to help you with that.  Selecting appropriate
>> packages is part of doing research, after all, and demonstrating
>> that you can do it is all to the good, no?
>>
>>
>> On Sun, 3 Jul 2022 at 15:24, Richard O'Keefe <raoknz using gmail.com> wrote:
>>
>>> Can we start a step back please?
>>> wmtsa stands for
>>>   Wavelet Methods for Time Series Analysis.
>>>
>>> OK, so you have some time series data,
>>> and for some reason you want to analyse
>>> your data using wavelets.  No worries.
>>> But does it have to be THIS unmaintained
>>> package?
>>>
>>> Why not visit
>>>
>>> https://CRAN.R-project.org/view=TimeSeries
>>> and search for "wavelets" in the text?
>>> Oh heck, I might as well do it for you.
>>> <snip>
>>> *Wavelet methods* : The wavelets
>>> <https://cran.r-project.org/web/packages/wavelets/index.html> package
>>> includes computing wavelet filters, wavelet transforms and multiresolution
>>> analyses. Multiresolution forecasting using wavelets is also implemented in
>>> mrf <https://cran.r-project.org/web/packages/mrf/index.html>.
>>> WaveletComp
>>> <https://cran.r-project.org/web/packages/WaveletComp/index.html>
>>> provides some tools for wavelet-based analysis of univariate and bivariate
>>> time series including cross-wavelets, phase-difference and significance
>>> tests. biwavelet
>>> <https://cran.r-project.org/web/packages/biwavelet/index.html> is a
>>> port of the WTC Matlab package for univariate and bivariate wavelet
>>> analyses. mvLSW
>>> <https://cran.r-project.org/web/packages/mvLSW/index.html> provides
>>> tools for multivariate locally stationary wavelet processes. LSWPlib
>>> <https://cran.r-project.org/web/packages/LSWPlib/index.html> contains
>>> functions for simulation and spectral estimation of locally stationary
>>> wavelet packet processes. Tests of white noise using wavelets are provided
>>> by hwwntest
>>> <https://cran.r-project.org/web/packages/hwwntest/index.html>. Wavelet
>>> scalogram tools are contained in wavScalogram
>>> <https://cran.r-project.org/web/packages/wavScalogram/index.html>.
>>> Further wavelet methods can be found in the packages rwt
>>> <https://cran.r-project.org/web/packages/rwt/index.html>, waveslim
>>> <https://cran.r-project.org/web/packages/waveslim/index.html>,
>>> wavethresh
>>> <https://cran.r-project.org/web/packages/wavethresh/index.html>.
>>> </snip>
>>>
>>> Presumably there is a reason that nobody else has
>>> bothered to continue maintaining wmtsa.  Perhaps
>>> one of those other wavelets + time series packages
>>> can do what you need?
>>>
>>>
>>> On Sun, 3 Jul 2022 at 04:12, Muhammad Zubair Chishti <
>>> mzchishti using eco.qau.edu.pk> wrote:
>>>
>>>> Dear Experts,
>>>> I cannot find a package "wmtsa" for my R version "R 4.2.0". Kindly help
>>>> me
>>>> to find it or share the link with me.
>>>> Although I tried the old version of "wmtsa" but failed.
>>>> Thank you for your precious time.
>>>>
>>>> Regards
>>>> Muhammad Zubair Chishti
>>>>
>>>>         [[alternative HTML version deleted]]
>>>>
>>>> ______________________________________________
>>>> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
>>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>>> PLEASE do read the posting guide
>>>> http://www.R-project.org/posting-guide.html
>>>> and provide commented, minimal, self-contained, reproducible code.
>>>>
>>>

	[[alternative HTML version deleted]]