Woods, D.C. and S.M. Lewis. "Continuous optimal designs for generalized linear models under model uncertainty." Journal of Statistical Theory and Practice. 5 (2011): 137-145. Electronic.
With more and more complex data and experiments, linear regression is often "inadequate." Even with the "modern" regression methods such as b-splines and smoothing kernels, standard factorial designs cannot be modeled well. This papers suggests are more exact designs for specific experiments using sophisticated design selection and criterion to allow uncertainty in the link (or knot) functions. The algorithm's efficiency is tested using simulation studies.
Again, this article helps me narrow down the statistics portion of my project. Understanding previous papers written by Dr. Woods helps me create a more solid literature review, helping me to develop my topic-specific information and significance.
Thursday, February 23, 2012
Wednesday, February 22, 2012
Source (2/22)
Woods, Dave and Peter van de Ven. "Blocked Designs for Experiments with Correlated Non-Normal Response." Technometrics. 53.2 (2011): 173-182. Electronic.
In simple linear models, the assumptions are very strict and often unattainable. Often, "many experiments measure a response that cannot be adequately described by a linear model with normally distributed errors." The authors developed a general method of creating efficient blocked designs where the response is distributed as an exponential family using Generalized Estimating Equations. "This methodology is appropriate when the blocking factor is a nuisance variable, as often occurs in industrial experiments." Using both a systematic search and a block optimal design for a Generalized Linear Model, the results are more efficient than using an optimal GLM design. This article is useful as I am trying to clarify my statistical part of the project. This allows me to gain an idea of the type of projects Dr. Woods is involved with.
In simple linear models, the assumptions are very strict and often unattainable. Often, "many experiments measure a response that cannot be adequately described by a linear model with normally distributed errors." The authors developed a general method of creating efficient blocked designs where the response is distributed as an exponential family using Generalized Estimating Equations. "This methodology is appropriate when the blocking factor is a nuisance variable, as often occurs in industrial experiments." Using both a systematic search and a block optimal design for a Generalized Linear Model, the results are more efficient than using an optimal GLM design. This article is useful as I am trying to clarify my statistical part of the project. This allows me to gain an idea of the type of projects Dr. Woods is involved with.
My Project, Take 2 (LJ 2/22)
I was just thinking how wonderful it is when you have no homework due the next day when I remembered I still have a non-stat class. And I have something due. Today. Probably a few hours ago. Oops.
As I was writing my Background, Significance, and Literature Review, I realized some unfortunate issues with my project. To begin with, I had a very, very rough time writing my draft. Though that may in part have been due to the fact I had just finished a grueling test, my paper should have gone more smoothly. Of course, I did not realize as I was writing that I was having issues with my project, not necessarily the writing. After the fact, I am now going to try to absolve my sins in my project.
I realized I have not yet clearly defined either part of my project, rendering the "specific information" very difficult to write. For the communication portion, I have not focused on what I want as the end goal nor on what I am defining as "academic communication". So: as a result of my project, I want to have strengthened the communications between Dr. Woods and BYU. In addition, I want to observe how professor-student communications differ in England as opposed to England. With those observed differences, I want to create a good connection between myself and Dr. Woods, an academic contact that I can continue to utilize in later years. As for the definition of "academic communication," I mean email contact, co-authored papers, weekly meetings, chance encounter conversations, clarifying questions, and presentations. There is probably more I will want to study, but the basic idea is "academic communication" is anything used to develop and sustain an academic relationship.
If possible, my statistics part of my project is even more vague. This is due in part to the fact that my actual project has not been decided. I only know the topic--design. But this in itself can mean anything from experimental design to computer simulations, practical to theoretical. Because I do not yet know what direction I will be pursuing, I cannot create a specific argument as to why I will be contributing to an "academic dialogue" nor can I give substantial "topic-specific information." As much as I would want to clarify my statistics project right now, I can't. At least not at the moment. I need to meet with Dr. Reese (my mentor) to iron out some details, and he may not even know what Dr. Woods wants to work on yet. I am kind of like the work force for part of this project. I just do whatever my professors tell me.
The most unfortunate part of my project is I am still trying to strengthen my argument that I need to do this project in England. Obviously I will have to if I am to work with Dr. Woods, but showing an "understanding of the local context" is a bit difficult. Honestly, I have not determined any "potential sensitivities specific to the culture and topic that may affect [my] field study." Maybe Dr. Woods will be unimpressed with my statistical background and send me back in disgrace? Maybe no one in Southampton will want to have me as an academic contact. Potentially. The project I will eventually work on may be too complex to solve in a summer, or ever. As it is the summer, I may have a difficult time finding students and professors to collaborate with at the university.
With the above clarifications now in place, I should probably re-write my part b. And I most certainly need to talk to Dr. Reese.
As I was writing my Background, Significance, and Literature Review, I realized some unfortunate issues with my project. To begin with, I had a very, very rough time writing my draft. Though that may in part have been due to the fact I had just finished a grueling test, my paper should have gone more smoothly. Of course, I did not realize as I was writing that I was having issues with my project, not necessarily the writing. After the fact, I am now going to try to absolve my sins in my project.
I realized I have not yet clearly defined either part of my project, rendering the "specific information" very difficult to write. For the communication portion, I have not focused on what I want as the end goal nor on what I am defining as "academic communication". So: as a result of my project, I want to have strengthened the communications between Dr. Woods and BYU. In addition, I want to observe how professor-student communications differ in England as opposed to England. With those observed differences, I want to create a good connection between myself and Dr. Woods, an academic contact that I can continue to utilize in later years. As for the definition of "academic communication," I mean email contact, co-authored papers, weekly meetings, chance encounter conversations, clarifying questions, and presentations. There is probably more I will want to study, but the basic idea is "academic communication" is anything used to develop and sustain an academic relationship.
If possible, my statistics part of my project is even more vague. This is due in part to the fact that my actual project has not been decided. I only know the topic--design. But this in itself can mean anything from experimental design to computer simulations, practical to theoretical. Because I do not yet know what direction I will be pursuing, I cannot create a specific argument as to why I will be contributing to an "academic dialogue" nor can I give substantial "topic-specific information." As much as I would want to clarify my statistics project right now, I can't. At least not at the moment. I need to meet with Dr. Reese (my mentor) to iron out some details, and he may not even know what Dr. Woods wants to work on yet. I am kind of like the work force for part of this project. I just do whatever my professors tell me.
The most unfortunate part of my project is I am still trying to strengthen my argument that I need to do this project in England. Obviously I will have to if I am to work with Dr. Woods, but showing an "understanding of the local context" is a bit difficult. Honestly, I have not determined any "potential sensitivities specific to the culture and topic that may affect [my] field study." Maybe Dr. Woods will be unimpressed with my statistical background and send me back in disgrace? Maybe no one in Southampton will want to have me as an academic contact. Potentially. The project I will eventually work on may be too complex to solve in a summer, or ever. As it is the summer, I may have a difficult time finding students and professors to collaborate with at the university.
With the above clarifications now in place, I should probably re-write my part b. And I most certainly need to talk to Dr. Reese.
Monday, February 20, 2012
Being Sensitive (LJ 2/21)
Though it may be slightly sacrilegious the way I always get my topic for my Monday learning journals from church, but there is such great material. Have you ever noticed the abundance of anthropological observations one can make in sacrament meeting alone? Then throw in Sunday School and Relief Society (I assume Elders' Quorum is similar, but I've never actually been), and you have an entire research field. I cannot count the number of times that I have been grateful that the National Geographic has not done an anthropological study about Mormons. Can you imagine having an outside anthropologist observing your Sunday School?
Anyway, sacrament meeting was all about the founding fathers because it is Presidents' Day weekend. I did not think anything of it. After all, it was nothing compared to my home ward's Four of July fast and testimony meeting. (Just so you can imagine: small Utahan town, chock full of redneck republicans. And the birthplace of David O. McKay. Very Mormon. Very patriotic.) And my poor roommate from Canada leans over and asks, "Is it normal to talk about the founding fathers in sacrament meeting?" Hmmmm. A little culturally insensitive? Because she is not an American and does not have extensive knowledge about our founding fathers, the topics alienated my roommate from sacrament meeting. And we are supposed to be a world religion.
Unfortunately, this cultural insensitivity did not register with me until someone else pointed it out to me. Frightening when you consider I am supposed to be preparing to be culturally sensitive. With this fact in mind, I decided to implement various aspects from class and common sense to purge me of my cultural insensitivities.
1. Think before I speak. Okay, this may be a generally good idea, not solely for the purpose of cultural sensitivity. But I need to make sure words coming out of my mouth will not betray unknown ignorance or bullheadedness. Did my bishopric even consider not everyone understands US history?
2. Be observant both here and in England. What defines US culture? English culture? Academic culture? For example, English statisticians use odds. We use probabilities. When I give a presentation in England, I need to make sure to use odds so I do not give myself away (because obviously my accent won't). And who knows what other differences there will be. Here, professors leave to go home at a variety of times. Is it more defined in Southampton? Professors joke with students. Is that taboo or accepted in England. So for now I need to understand what defines the BYU statistics department culture for a stronger comparison.
3. Realize and then ignore predetermined ideas. Though I have not noticed yet, I may have some strong prejudices. I don't think the US is the most amazing country in the world, but there may be something else blisteringly obvious. Maybe like I think the class system in England is ridiculous. Or that Utah is the best place in the world. Little ideas like those. Once recognized, I'll be able to observe the actuality of the ideas.
Unfortunately, this cultural insensitivity did not register with me until someone else pointed it out to me. Frightening when you consider I am supposed to be preparing to be culturally sensitive. With this fact in mind, I decided to implement various aspects from class and common sense to purge me of my cultural insensitivities.
1. Think before I speak. Okay, this may be a generally good idea, not solely for the purpose of cultural sensitivity. But I need to make sure words coming out of my mouth will not betray unknown ignorance or bullheadedness. Did my bishopric even consider not everyone understands US history?
2. Be observant both here and in England. What defines US culture? English culture? Academic culture? For example, English statisticians use odds. We use probabilities. When I give a presentation in England, I need to make sure to use odds so I do not give myself away (because obviously my accent won't). And who knows what other differences there will be. Here, professors leave to go home at a variety of times. Is it more defined in Southampton? Professors joke with students. Is that taboo or accepted in England. So for now I need to understand what defines the BYU statistics department culture for a stronger comparison.
3. Realize and then ignore predetermined ideas. Though I have not noticed yet, I may have some strong prejudices. I don't think the US is the most amazing country in the world, but there may be something else blisteringly obvious. Maybe like I think the class system in England is ridiculous. Or that Utah is the best place in the world. Little ideas like those. Once recognized, I'll be able to observe the actuality of the ideas.
Source (2/21)
Rencher, Alvin, Bruce Schaalje. Linear Models in Statistics. Hoboken: John Wiley & Sons, Inc. 2008. Print.
Linear models are the basis for just about anything statistics related. Having a solid foundation in regression theory and application will help me understand and extend the theory of design studies. In addition to developing linear regression models, this book has a chapter specifically devoted to design theory for both Analysis of Variance models and Analysis of Covariance models. Also, this book explains the theory behind hypothesis testing and what to do when certain assumptions are violated in an experiment. In short, it explains both how to use statistics to design an experiment and how to redeem a poorly designed experiment. Understanding the principles in this book will help me progress with the theory and application behind the statistics in my project.
Linear models are the basis for just about anything statistics related. Having a solid foundation in regression theory and application will help me understand and extend the theory of design studies. In addition to developing linear regression models, this book has a chapter specifically devoted to design theory for both Analysis of Variance models and Analysis of Covariance models. Also, this book explains the theory behind hypothesis testing and what to do when certain assumptions are violated in an experiment. In short, it explains both how to use statistics to design an experiment and how to redeem a poorly designed experiment. Understanding the principles in this book will help me progress with the theory and application behind the statistics in my project.
Friday, February 17, 2012
My nemesis: Simulations (LJ 2/17)
I took my first computer programming class in the winter of my discontent, 2010. We learned (or should have learned) Java, and I absolutely hated it. Not only did I decide to never, ever choose another class or career potentially related to programming, I almost decide to ditch the math major. Barely scraping through the class, I threw in the towel as soon as my final was sort of completed. That summer, I got a fellowship, but I did not read the fine print. Or the full title. It was Interdisciplinary Mentoring Program in Analysis, Computation, and Theory. Yep, I did programming all summer. Once I completed the training, I breathed a sigh of relief: for sure I was done with programming, and I could get back to theoretical math. Nope--my mentor was a statistics professor, and my project dealt mainly with creating a computer function that modeled certain RNA strands in breast cancers.
Since becoming a grad student in statistics, I have learned, with much griping and grumbling, to accept that programming is here to stay and dominate most of my statistical life. So I vanquished my archenemy. But like Sherlock, I have to have a nemesis. While I enjoy most aspects of programming, I struggle with simulation studies. Essentially, a simulation study involves generating a lot of data given various parameters. They are very useful when collecting a lot of "real" data is expensive, unethical, or impossible. For example, atmospheric data, cancer data, etc. Useful yes, easy to accomplish--not at all. My new archenemy. I only got through my first simulation study because I thought it would be my last.
Given simulations haunt me continually, you can guess my horror when I realized what my project entails. Simulation studies. This really hit me yesterday. The statistics department had Dr. Finley as a guest speaker who elaborated on exactly what a computer experiment was. I do not know why I did not connect the terms and think what they meant, but I finally realized computer experiments is just another term for simulation studies.
However after listening to Dr. Finley explain about his research, I had a similar epiphany that I had after listening to the seminar by Dr. Heaton. Studying simulations (especially in London) will help me understand them more so I may apply them to different fields (i.e forestry like Dr. Finley or the atmosphere like Dr. Heaton). Though still a little apprehensive about struggling through the simulations, I am excited that this is my project for England. It is helping me to accept my homework and pay more attention as this is what I'll be doing forever, specifically over the summer.
Since becoming a grad student in statistics, I have learned, with much griping and grumbling, to accept that programming is here to stay and dominate most of my statistical life. So I vanquished my archenemy. But like Sherlock, I have to have a nemesis. While I enjoy most aspects of programming, I struggle with simulation studies. Essentially, a simulation study involves generating a lot of data given various parameters. They are very useful when collecting a lot of "real" data is expensive, unethical, or impossible. For example, atmospheric data, cancer data, etc. Useful yes, easy to accomplish--not at all. My new archenemy. I only got through my first simulation study because I thought it would be my last.
Given simulations haunt me continually, you can guess my horror when I realized what my project entails. Simulation studies. This really hit me yesterday. The statistics department had Dr. Finley as a guest speaker who elaborated on exactly what a computer experiment was. I do not know why I did not connect the terms and think what they meant, but I finally realized computer experiments is just another term for simulation studies.
However after listening to Dr. Finley explain about his research, I had a similar epiphany that I had after listening to the seminar by Dr. Heaton. Studying simulations (especially in London) will help me understand them more so I may apply them to different fields (i.e forestry like Dr. Finley or the atmosphere like Dr. Heaton). Though still a little apprehensive about struggling through the simulations, I am excited that this is my project for England. It is helping me to accept my homework and pay more attention as this is what I'll be doing forever, specifically over the summer.
Wednesday, February 15, 2012
Oh Ophelia (LJ 2/15)
Poor Ophelia. She maybe is a little naive and "chronically submissive," but is it her fault? Does she deserve to be the champion of a rather unpleasant syndrome? I feel the blame lies with Polonius for brainwashing Ophelia from the beginning. In addition to disagreeing with the name, Thomas Plummer's "Diagnosing and Treating the Ophelia Syndrome" was almost a catch-22: my initial reaction to my reading was to accept it as "the TRUTH". There are no more treatments and those listed are of course perfectly suited to everyone--the author then becomes a Polonius. But Plummer urges us to "develop a healthy distrust of authorities and experts" which includes himself. So we should distrust him and not question authorities? Very mind-boggling.Moving past the catch-22, Plummer's treatments have potential both for curing or creating an epic failure. For example, some may take treatment 3, learning to live with uncertainty, as a cop out to trying to understand. After looking at a problem for hours, I am willing to accept defeat and never mind I should be able to understand. I could just take treatment 3 and live with never trying harder.
However, as in all treatments, moderation and critical thinking (and applying) is key. One treatment I am anxious to implement in London is stepping out of bounds in London. Don't panic; mainly I only mean breaking self-imposed bounds. Last summer, I had an internship in Boston. And I watched a lot of Dr. Who episodes. A lot a lot. I barely infiltrated the city. My bounds told me to avoid Boston at night alone, and my fears told me to avoid people in general. I allowed my fear and pointless discretion dictate what I could do. Not to be repeated in London. By stepping out of my bounds, I will be able to observe and learn more about London and myself. Time to stop inflicting rules on myself and holding myself back (within reason). London has so many opportunities, but I have to be willing to break the bounds to have them. If I never question authority or always meekly accept archaic rules imposed by myself to avoid getting hurt or lost, I will not be able to really push the boundaries of statistics, truly establish an academic contact, or honestly observe people and myself.
Subscribe to:
Posts (Atom)