Comments on: Categories http://hea-www.harvard.edu/AstroStat/slog/2007/categories/ Weaving together Astronomy+Statistics+Computer Science+Engineering+Intrumentation, far beyond the growing borders Fri, 01 Jun 2012 18:47:52 +0000 hourly 1 http://wordpress.org/?v=3.4 By: hlee http://hea-www.harvard.edu/AstroStat/slog/2007/categories/comment-page-1/#comment-13 hlee Mon, 28 May 2007 02:29:30 +0000 http://hea-www.harvard.edu/AstroStat/slog/2007/categories/#comment-13 I didn't imply only heteroscedasticity nor the distinction between standard error and standard deviation. These are a few statisticians' errors. I think measurement errors and misidentifications are modeled in statistics but not considered in astronomy. I wish i could explain more statisticians' errors but beyond my limit. [<strong>Response:</strong> I don't think Astronomers think of errors in a particularly different manner than statisticians. We generally don't think like frequentists though, so we have measurement errors and model parameter error bars. ] Anyhow, I almost returned all my shuffles in (sub)categories to your original proposal with a few additions as you and Alanna suggested. My minor worry is that statisticians may feel awkward to find categories as you felt strangeness when I changed the structure of categories. I wish that experienced statisticians comment and make suggestions. [<strong>Response:</strong> If you feel that some subcategory does not belong under "<code>Stat</code>", then simply move it out to its own category. Please also move "<code>Uncertainty</code>" out from under "<code>Astro</code>" -- it most definitely does not belong there. Also, I think a couple more are missing, such as "<code>Meta</code>" and "<code>Bad AstroStat</code>". Also, I think "<code>Astro:Objects</code>" is too vague and could be eliminated. ] One thing strucks me. Add another posting explains what topics are suitable or what researches CHASC has been done for these categories. This will be helpful who access our slog and useful to be added at "About" or "Disclaimer" (I haven't done the latter yet due to lack of proper wordings") I didn’t imply only heteroscedasticity nor the distinction between standard error and standard deviation. These are a few statisticians’ errors. I think measurement errors and misidentifications are modeled in statistics but not considered in astronomy. I wish i could explain more statisticians’ errors but beyond my limit.

[Response: I don't think Astronomers think of errors in a particularly different manner than statisticians. We generally don't think like frequentists though, so we have measurement errors and model parameter error bars. ]

Anyhow, I almost returned all my shuffles in (sub)categories to your original proposal with a few additions as you and Alanna suggested. My minor worry is that statisticians may feel awkward to find categories as you felt strangeness when I changed the structure of categories. I wish that experienced statisticians comment and make suggestions.

[Response: If you feel that some subcategory does not belong under "Stat", then simply move it out to its own category. Please also move "Uncertainty" out from under "Astro" -- it most definitely does not belong there. Also, I think a couple more are missing, such as "Meta" and "Bad AstroStat". Also, I think "Astro:Objects" is too vague and could be eliminated. ]

One thing strucks me. Add another posting explains what topics are suitable or what researches CHASC has been done for these categories. This will be helpful who access our slog and useful to be added at “About” or “Disclaimer” (I haven’t done the latter yet due to lack of proper wordings”)

]]>
By: HLee http://hea-www.harvard.edu/AstroStat/slog/2007/categories/comment-page-1/#comment-11 HLee Fri, 25 May 2007 19:40:54 +0000 http://hea-www.harvard.edu/AstroStat/slog/2007/categories/#comment-11 <p>Statisticians have different "errors" "uncertainties" and some others. Because of we have astronomical errors/data, we could find such statistical methods that you've described. Same statistical methods can stand alone with targeting sociology data without errors but repeated measures so that statisticians may get confused of such subcategorizing. </p> <p>Also, as you aware, one can mark as many categories and subcategories with their posting. I was only afraid of too many categories with complicated hierarchical structure (like BLoCXS). I prefer to separating subcategories based on their origins. </p> <p>[<strong>Response:</strong> While that is true, the posts get organized in cascading subcategories, and it will be jarring to find a post on error bars under <code>Astro</code>, or <code>wavdetect</code> under <code>Stat:Methods</code>, etc. If there is a chance that a given subcategory spans multiple disciplines, keep it distinct. -vlk ]</p> <p>In that regards, I like Alanna's suggestion (and I will replace Algorithms to Methods)<br /> Stat>Methods>MCMC<br /> >Fitting (those algorithms)<br /> >Important Sampling Bootstrap [<strong>Response:</strong> Do not replace <code>Algorithms</code>. Not all algorithms are statistical. Morphological analyses, for example, won't fit under this scheme. <code>Algorithms</code> and <code>Stat:Methods</code> need to be separate categories. -vlk]</p> <p>I'll reflect this changes tomorrow. I'm not thinking clearly now. </p> <p><i> By the way, astronomers' "stat error" is different from statisticians' "error" There was a book about errors written by statisticians, which I cannot recall. </i></p> <p>[<strong>Response:</strong> I think this is simply the distinction between standard error and standard deviation, and the nuances that come in due to heteroscedasticity. -vlk]</p> Statisticians have different “errors” “uncertainties” and some others. Because of we have astronomical errors/data, we could find such statistical methods that you’ve described. Same statistical methods can stand alone with targeting sociology data without errors but repeated measures so that statisticians may get confused of such subcategorizing.

Also, as you aware, one can mark as many categories and subcategories with their posting. I was only afraid of too many categories with complicated hierarchical structure (like BLoCXS). I prefer to separating subcategories based on their origins.

[Response: While that is true, the posts get organized in cascading subcategories, and it will be jarring to find a post on error bars under Astro, or wavdetect under Stat:Methods, etc. If there is a chance that a given subcategory spans multiple disciplines, keep it distinct. -vlk ]

In that regards, I like Alanna’s suggestion (and I will replace Algorithms to Methods)
Stat>Methods>MCMC
>Fitting (those algorithms)
>Important Sampling Bootstrap [Response: Do not replace Algorithms. Not all algorithms are statistical. Morphological analyses, for example, won't fit under this scheme. Algorithms and Stat:Methods need to be separate categories. -vlk]

I’ll reflect this changes tomorrow. I’m not thinking clearly now.

By the way, astronomers’ “stat error” is different from statisticians’ “error” There was a book about errors written by statisticians, which I cannot recall.

[Response: I think this is simply the distinction between standard error and standard deviation, and the nuances that come in due to heteroscedasticity. -vlk]

]]>
By: HLee http://hea-www.harvard.edu/AstroStat/slog/2007/categories/comment-page-1/#comment-10 HLee Fri, 25 May 2007 17:32:30 +0000 http://hea-www.harvard.edu/AstroStat/slog/2007/categories/#comment-10 <p>From an engineer point of view, imaging, spectral, timing can be considered as data processing (my small opinion due to my minor degree in EE :)).</p> <p>[<strong>Response:</strong> It could be under that category, but not limited to it. With Alanna's blob significance method, it could be under stat errors, for example. I think the point of these categories is to be as general and flexible as possible. We don't know what we are going to write about two months from now, and don't want to be scrambling to rearrange things then.</p> <p>PS: As Alanna suggested, another useful category is <code><strong>Stat > Methods</strong></code></p> <p>-vlk]</p> From an engineer point of view, imaging, spectral, timing can be considered as data processing (my small opinion due to my minor degree in EE :) ).

[Response: It could be under that category, but not limited to it. With Alanna's blob significance method, it could be under stat errors, for example. I think the point of these categories is to be as general and flexible as possible. We don't know what we are going to write about two months from now, and don't want to be scrambling to rearrange things then.

PS: As Alanna suggested, another useful category is Stat > Methods

-vlk]

]]>
By: Hyunsook http://hea-www.harvard.edu/AstroStat/slog/2007/categories/comment-page-1/#comment-9 Hyunsook Fri, 25 May 2007 16:52:29 +0000 http://hea-www.harvard.edu/AstroStat/slog/2007/categories/#comment-9 Probably, I saw your posting to early. 1969 seems default. Some changes made. Meta (about the site, the software, etc) [<strong>Response:</strong> Make that "<code>admin</code>" or something like that then. ] Everything automatically become alphabetical. <strong>Algorithms Algorithms>Fitting Algorithms>MCMC [<strong>Response:</strong> I think this is too restrictive. We don't want to speak of fitting and MCMC <em>only</em> in the context of <code>algorithms</code>. They really belong under <code>Stat</code>. ] <strong>HLee:</strong> The MCMC workshop by Jun Liu proved that algorithms does not only belong to Stat. Algorithms can stand alone. Maybe change the name, <code>Computation</code> Astro Astro > High-Energy Astro > High-Energy > gamma-ray Astro > High-Energy > X-ray Astro > Imaging Astro > Objects Astro > Objects > Galaxies Astro > Objects > Stars Astro > Objects > Sun [<strong>Response:</strong> The Sun is a star. It should be <code>Astro > Objects > Stars > Sun</code> in this case. Having "<code>Objects</code>" seems redundant. "<code>Other</code>" may be more relevant. ] <strong>HLee:</strong> Two different subjects. Solar Physics and Stellar Astronomy. I get rid of Sun since you didn't put it your proposal. I just put it because of some solar persons.</b> Astro > Optical Astro > Physics Astro > Spectral Astro > Timing [<strong>Response:</strong> <code>Imaging</code>, <code>Spectral</code>, and <code>Timing</code> should not be limited to "<code>Astro</code>". ] <strong>HLee:</strong> These are subjects of Signal/Image Processing (Timing might be an interest in survival analysis?), not direct objectives of statistics. You want to make them stand alone? I thought there are too many parent categories. I'll move them to the top. CHASC Data Processing Jargon Languages News Nuggets Quotes Stat Stat>Bayesian Stat>Frequentist</strong> I intentionally omit "error" because we deal astronomical errors with statistical methods. MCMC move to algorithms since MC was already popular with scientists/engineers and MCMC seems not only belong to statistics. Fitting was moved along the same reasoning. [<strong>Response:</strong> Make that "<code>uncertainty</code>" then. ] <strong>HLee:</strong> Under Astro [<strong>Response:</strong> No no, not under Astro! Errors may mean error bars, confidence levels, credible regions, confidence bounds, detection limits, FDR, Type I, Type II, etc. NOT under Astro. -vlk] Changes require for "Stat" to have more subcategories, I'll add them. Like nonparametric, machine learning (PCA, classification/clustering stuffs) By the way, all these categories were not visible until someone post under the given categories, I guess. <i>HLee: I haven't found how to upload plug-ins under /wp-contents/plug-in at groundtruth.info. We need inline response to make commenting easy.. (sweat) </i> [<strong>Response:</strong> I am using a construct of the form "<p class="response">[<strong>Response:</strong> xxx ]</p>", as suggested by Gavin of RC. The css can be changed to include something about this class, and a javascript button can be added to the edit page to automatically add that snippet. -vlk] Probably, I saw your posting to early. 1969 seems default.
Some changes made.
Meta (about the site, the software, etc)
[Response: Make that "admin" or something like that then. ]
Everything automatically become alphabetical.
Algorithms
Algorithms>Fitting
Algorithms>MCMC
[
Response: I think this is too restrictive. We don't want to speak of fitting and MCMC only in the context of algorithms. They really belong under Stat. ]
HLee: The MCMC workshop by Jun Liu proved that algorithms does not only belong to Stat. Algorithms can stand alone. Maybe change the name, Computation
Astro
Astro > High-Energy
Astro > High-Energy > gamma-ray
Astro > High-Energy > X-ray
Astro > Imaging
Astro > Objects
Astro > Objects > Galaxies
Astro > Objects > Stars
Astro > Objects > Sun
[Response: The Sun is a star. It should be Astro > Objects > Stars > Sun in this case. Having "Objects" seems redundant. "Other" may be more relevant. ]
HLee: Two different subjects. Solar Physics and Stellar Astronomy. I get rid of Sun since you didn’t put it your proposal. I just put it because of some solar persons.
Astro > Optical
Astro > Physics
Astro > Spectral
Astro > Timing
[Response: Imaging, Spectral, and Timing should not be limited to "Astro". ]
HLee: These are subjects of Signal/Image Processing (Timing might be an interest in survival analysis?), not direct objectives of statistics. You want to make them stand alone? I thought there are too many parent categories. I’ll move them to the top.
CHASC
Data Processing
Jargon
Languages
News
Nuggets
Quotes
Stat
Stat>Bayesian
Stat>Frequentist
I intentionally omit “error” because we deal astronomical errors with statistical methods. MCMC move to algorithms since MC was already popular with scientists/engineers and MCMC seems not only belong to statistics. Fitting was moved along the same reasoning.
[Response: Make that "uncertainty" then. ]
HLee: Under Astro

[Response: No no, not under Astro! Errors may mean error bars, confidence levels, credible regions, confidence bounds, detection limits, FDR, Type I, Type II, etc. NOT under Astro. -vlk]

Changes require for “Stat” to have more subcategories, I’ll add them. Like nonparametric, machine learning (PCA, classification/clustering stuffs)
By the way, all these categories were not visible until someone post under the given categories, I guess.
HLee: I haven’t found how to upload plug-ins under /wp-contents/plug-in at groundtruth.info. We need inline response to make commenting easy.. (sweat)

[Response: I am using a construct of the form "<p class="response">[<strong>Response:</strong> xxx ]</p>”, as suggested by Gavin of RC. The css can be changed to include something about this class, and a javascript button can be added to the edit page to automatically add that snippet. -vlk]

]]>
By: Hyunsook http://hea-www.harvard.edu/AstroStat/slog/2007/categories/comment-page-1/#comment-8 Hyunsook Fri, 25 May 2007 16:15:29 +0000 http://hea-www.harvard.edu/AstroStat/slog/2007/categories/#comment-8 I'll see how the child categories actually can be seen. I'll work on it. By the way, why all your draft postings were dated 1969? [<strong>Response:</strong> I have no idea! -vlk] I’ll see how the child categories actually can be seen. I’ll work on it.

By the way, why all your draft postings were dated 1969?

[Response: I have no idea! -vlk]

]]>