Posts by rosemary7962

rosemary7962

Oct 19, 2015

Graduate / Economics Ph.D. SOP - need advice for expunging redundant information [7]

@vangiespen :D I don't have any working experiences. I just graduated form master's this June; and I pursued my M.A. soon after I graduated form B.A. If there is any working experience, it is what I am now doing: a research assistant. I decided to be a research assistant because I want to prepare well for the following PhD program, so that I would know what is on the frontier of economics researches.

May I explain the situation of the academia of Economics? Most of the people who decided to pursue a PhD would enroll in a Master's just after their Bachelor's, and apply for a PhD as soon as possible after they got their Master's. (This phenomena is revealed in the CVs of the faculty members and PhD candidates of most of the programs I plan to apply for, as well as all the professors I know in my country.) During the years of their pre-doctoral education, they would take as many math related courses as they can because economics now is kind of similar to mathematics (though mathematicians won't agree, ha).

I understand that a sop should be career oriented; however, the only job I have ever got is a RA. If what invokes me to pursue a PhD is researching itself, then how would you advise me to organize my sop?

A million thanks :D

rosemary7962

Oct 19, 2015

Graduate / Economics Ph.D. SOP - need advice for expunging redundant information [7]

@vangiespen, thanks for your advice! I wrote a new version and I tried to focus on my future plan. Yet, I have a question. I never wrote this kind of essay before since it is not required for applying to a university in my country. I read the instructions on the websites of the programs I plan to apply, and the websites say that the SOP is limited to 1000 words. I am a little confused: should I trim the SOP as short as possible?

Thank you.
--------------------------------------------------

When I was a junior in National Taiwan University (NTU), the best university in Taiwan, I attended NTU Public Policy Forum, which discussed the asymmetric intervention of Taiwan's central bank in foreign exchange market. In the forum, several papers employed time series analysis to discuss how central bank behaves after New Taiwanese Dollar fluctuates. I suddenly realized how data analysis matters in explaining a social phenomena and my desire to become a econometrician is sparked.

To equip myself with knowledge required to become a independent researcher, I put a lot of efforts into different kinds of self-training, such as taking math related courses, honing programming skills as well as refining researching ability. During my Bachelor's, I have already taken mathematics courses such as Advanced Calculus (convergence theories) and Linear Algebra. Realizing how math helps me in economics analysis, it inspired me to take Real Analysis (measure theory), Advanced Statistical Inference, (Measure-Theoretic) Probability Theory as well as Stochastic Calculus (Brownian motion and Ito calculus) in my Master's. I also learned C language in my Bachelor's; this rigorous training in coding helps me a lot in my later research experience with statistical programs like MATLAB and R: I become a lecturer of text mining and a GitBook coeditor in R-Ladies, a branch of Taiwan R User's Group, and have being R programming TA in doctoral level Econometrics Theory class for three years. My well prepared math and programming skills allowed me to replace my Master's core curriculum with Ph.D. core curriculum and allows me to become a TA of graduate level Econometrics Theory. In the Ph.D. Econometric Theory Class, there I met my thesis advisor, Prof. AAA.

My M.S. thesis intend to use large dimension of macroeconomic variables to forecast DGP growth rate by supervised factor (SF) model. SF model in my thesis includes partial least square, principal covariate regression and combining forecast principal component analysis. These methods, comparing to dynamic factor (DF) model, premixed the variable of interest with the predictors before factorization process, thus the factors have better forcastability than DF model. Because of the good quality of both analytical and empirical part in my M.S. thesis, it is honored with Outstanding Master's Thesis Award among national-wide competitors and is presented in 2014 Taiwan Econometrics Society Annual Conference and NTU econometrics seminar.

My research interest expands with the development of my M.S. thesis. In particular, I am not only interested in economic forecast and applied macroeconomics, but also fascinated by econometrics theory topics like model selection (including variable selection and GMM moment selection), mixed data frequency, and Bayesian econometrics. I studied variable selection methods like least absolute shrinkage and selection operator (LASSO), least angle regression (LARS) as well as orthogonal greedy algorithm (OGA) and mixed data frequency (MIDAS) method proposed in Ghysels, Sinko, and Valkanov (2007) in the empirical part of my thesis. Variable selection invoked my further independent study of GMM moment selection problem (advised by Prof. BBB) later. There is an adroit method of GMM moment selection by LASSO proposed in Liao (2013); however, methods like LARS and OGA is not applied in GMM moment selection. Mixed data frequency and calendar effect problem are also topics that I want to improve. I have employed MIDAS method in my thesis while the forecasting result is confusingly bad. I think GMM moment selection and mixed data frequency in forecasting are possible topics for my Ph.D. study.

The predictability of reduced form model, comparing to structural form macroeconomic model, is my another concern. I am interested in asymmetric intervention of central bank in foreign exchange market, so I did a independent research (advised by Prof. CCC) to estimate the parameters in small open economic DSGE model with Taiwan's macroeconomic data by Bayesian MCMC method. The result reveals asymmetric distributed coefficient for foreign exchange rate in Taylor rule; however, the dynamics is not available. Time-varying parameters model is popular in time series and is possible to be applied to structural model for predicting central bank's behavior. This is another topic I would focus on during my Ph.D.

To become a top researcher, I always try to catch up with the mainstream. I work as a research assistant now under Prof. AAA and is surveying the three approaches to control endogeneity, Instrumental Variable, Control Function ans Fitted value, both their parametric and non-parametric versions. We plan to gave brief instructions to let laymen follow to deal with the endogeneity in their data and do an empirical study in corporate finance as a example. I think non-parametric methods is a promising approach to many questions and I am trying hard to master it: I plan to audit Topics on Functional Analysis (I) in spring 2016 in NTU.

(The last paragraph would vary with the faculty of different programs)

rosemary7962

Oct 15, 2015

Graduate / Economics Ph.D. SOP - need advice for expunging redundant information [7]

Hi, I am writing my SOP for applying Economics PhD. I have finish the part of my previous works and my research interest, and found that it is already about 800 words now. I plan to keep this sop in about 1000 words, while I still have my TA (of graduate level core curriculum), RA (what I am now researching), extracurricular activity (gave a text mining lecture in Taiwan R User Group) experiences, and of course linkage my research interest to the faculty of each school need to include.

Can you give me some advice to remove the redundant information as well as where can I put the extra experience listed above? Any comment is very welcome and thank you in advance.

Note: Prof. AAA, Prof. BBB, Prof. CCC, Prof. DDD would write me my recommendation letters.

-------------------------------------------------------------------------------
It was in National Taiwan University (NTU) Public Policy Forum that I firstly touched the issue of asymmetric intervention on foreign exchange market of central bank. In pursuing the existence of asymmetric intervention on foreign exchange market, it eventually bends my path to econometric research. When I was a junior in NTU, the best university in Taiwan, I took a course of Monetary and Banking and started to gain interest in monetary policy. Since I have taken some mathematics courses like Advanced calculus (Introductory mathematical analysis) and linear algebra, I prefer more rigorous approaches than just having intuition: I began to attend different kinds of seminars to learn how economists analyze economic issues. Once, NTU Public Policy Forum was held to discuss the influence of monetary policy on economy by time series analysis, and I was fascinated.

In the forum, several papers employ time series analysis to discuss how central bank behaves after New Taiwanese Dollar fluctuates, and I suddenly found how data analysis matters in researching. So later in my first year of Master's degree, I took a course of time-series-approach international finance, which taught by Prof. AAA, and write a term paper to evaluate how foreign direct investment (FDI) influence gross domestic product (GDP) through the financial market channel. I use several benchmarks like the credit of private sector and stock market capitalization to evaluate the efficiency of financial market and employed basic panel data model to estimate the model. I learned that even though the theory about this issue is still in discuss, data do confirm us the existence of some effect. This thought corroborates my resolution of pursuing a Ph.D.

In order to analyze data properly, I took several mathematics courses such as real analysis (measure theory), advanced statistical inference, (measure-theoretic) probability theory as well as stochastic calculus during my Master's degree. I also took doctoral level econometric theory taught by Prof. BBB to learn rigorous theory of econometrics. Because my interest in time series analysis meets Prof. BBB's research field of interest, my M.S. thesis is advised by Prof. BBB. My M.S. thesis intend to use large dimension of macroeconomic variables to forecast DGP growth rate by supervised factor (SF) model. SF model in my thesis includes partial least square, principal covariate regression and combining forecast principal component analysis. These methods, comparing to dynamic factor (DF) model, premixed the variable of interest with the macroeconomic variables before factorization process, thus the factors have better predictability than DF model. In order to enhance forecast performance, I also consider to preselect the macroeconomic variables by variable selection methods such as least absolute shrinkage and selection operator (LASSO), least angle regression (LARS) as well as orthogonal greedy algorithm (OGA). They invoked my further study later despite variable selection is not considered in my thesis in the end because it will be clearer, if I employed the same dataset as other previous works of DF model, to focus on the improvement of predictability of SF models on DF model. I think research about SF model is promising because several topics such as variable preselection and mixed data frequency (which I also considered during the empirical part) can be applied to improve the forecast. My M.S. thesis is honored with Outstanding Master's Thesis Award given by Taiwan Economic Society and is presented in 2014 Taiwan Econometrics Society Annual Conference and NTU econometrics seminar.

My experience of writing M.S. thesis led me to explore further topics such as whether the variable selection method can be applied to moment selection problem under generalized method of moment (GMM). I did a independent study about this topic advised by Prof. CCC. Meanwhile, I am still learning other methods such as Bayesian estimation in Prof. DDDFpro's Applied Econometric Method class and wrote a term paper trying to look at the asymmetric intervention of Taiwan's foreign exchange market by Bayesian Markov chain Monte Carlo method to estimate the parameter of a small open economy dynamic stochastic general equilibrium (DSGE) model. In these term papers, I encountered some problems that could be research during my Ph.D. For instance, the variable selection methods such as LARS and OGA rank the variables by the variable of interest in regression model, while under GMM scheme, there is no variable of interest; if any, it is zero. On the other hand, the estimation of DSGE model by Bayesian method would have firstly transform the model into state-transition equations and then use Kalman filter to get the prediction. The advantage of Bayesian estimation is its data driven property; however, Kalman filter is based on normal distribution, which makes the estimation not that data-driving.

(The next paragraph should link my interest to faculty member's research.)