Econometrics - Science topic

What is this Question refering to

The website Eventstudytools provides tools and resources to conduct event studies, including the BHAR (Buy-and-Hold Abnormal Return) approach. The BHAR approach is commonly used in event studies to calculate abnormal returns over a specific event period.

Eventstudytools is a comprehensive platform that offers a range of functionalities to analyze the impact of events on financial markets. It provides pre-built event study templates, customizable event windows, and statistical tests to measure abnormal returns. These tools can help researchers and analysts assess the significance of events and their impact on stock prices or other financial variables.

I would like to restart this question.

I have noted a few papers on cutoff or quasi-cutoff sampling other than the many I have written, but in general, I do not think those others have had much application. Further, it may be common to ignore the part of the finite population which is not covered, and to only consider the coverage, but I do not see that as satisfactory, so I would like to concentrate on those doing inference. I found one such paper by Guadarrama, Molina, and Tillé which I will mention later below.

Following is a tutorial i wrote on quasi-cutoff (multiple item survey) sampling with ratio modeling for inference, which can be highly useful for repeated official establishment surveys:

"Application of Efficient Sampling with Prediction for Skewed Data," JSM 2022: 

https://www.researchgate.net/publication/362370770_Application_of_Efficient_Sampling_with_Prediction_for_Skewed_Data 

This is what I did for the US Energy Information Administration (EIA) where I led application of this methodology to various establishment surveys which still produce perhaps tens of thousands of aggregate inferences or more each year from monthly and/or weekly quasi-cutoff sample surveys. This also helped in data editing where data collected in the wrong units or provided to the EIA from the wrong files often showed early in the data processing. Various members of the energy data user community have eagerly consumed this information and analyzed it for many years. (You might find the addenda nonfiction short stories to be amusing.)

There is a section in the above paper on an article by Guadarrama, Molina, and Tillé(2020) in Survey Methodology, "Small area estimation methods under cut-off sampling," which might be of interest, where they found that regression modeling appears to perform better than calibration, looking at small domains, for cutoff sampling. Their article, which I recommend in general, is referenced and linked in my paper.

There are researchers looking into inference from nonprobability sampling cases which are not so well-behaved as what I did for the EIA, where multiple covariates may be needed for pseudo-weights, or for modeling, or both. (See Valliant, R.(2019)*.) But when many covariates are needed for modeling, I think the chances of a good result are greatly diminished. (For multiple regression, from an article I wrote, one might not see heteroscedasticity that should theoretically appear, which I attribute to the difficulty in forming a good predicted-y 'formula'. For psuedo-inclusion probabilities, if many covariates are needed, I suspect it may be hard to do this well either, but perhaps that may be more hopeful. However, in Brewer, K.R.W.(2013)**, he noted an early case where failure using what appears to be an early version of that helped convince people that probability sampling was a must.)

At any rate, there is research on inference from nonprobability sampling which would generally be far less accurate than what I led development for at the EIA.

So, the US Energy Information Administration makes a great deal of use of quasi-cutoff sampling with prediction, and I believe other agencies could make good use of this too, but in all my many years of experience and study/exploration, I have not seen much evidence of such applications elsewhere. If you do, please respond to this discussion.

Thank you - Jim Knaub

..........

*Valliant, R.(2019), "Comparing Alternatives for Estimation from Nonprobability Samples," Journal of Survey Statistics and Methodology, Volume 8, Issue 2, April 2020, Pages 231–263, preprint at 

https://www.researchgate.net/publication/337950671_Comparing_Alternatives_for_Estimation_from_Nonprobability_Samples 

**Brewer, K.R.W.(2013), "Three controversies in the history of survey sampling," Survey Methodology, Dec 2013 -  Ken Brewer - Waksberg Award article: 

It is merely a scale-localized version of the usual cross-correlation between two signals. In cross-correlation, you determine the similarity between two sequences by shifting one relative to the other, multiplying the shifted sequences element by element, and summing the result.

Wavelet cannot be used to test for cointegration

Robustness measures the degree of invariance of some feature F across some class of entities E. Sensitivity analyses are forms of robustness measurement (F = model outcome, E = model inputs/parameters/structure variations).

Hi Aya Galal ,

Eleven (11) variables for a small sample period of 27 years (n = 27) are considerably too much and will lead to over-specification of the model.

Based on your research objective(s), I suggest you please revisit your economic theories regarding the relationships your chosen dependent variable has with each of the regressors/explanatory variables, and choose those with highest relevance in line with the underlying theories to be tested.

If the set or number of explanatory variables remains large, I suggest you perform some data reduction of such variables using Principal Component Analysis (PCA) approach, and retain the independent variables that have highest loadings across components - you can use your expert judgement to determine the threshold cut-off of the loadings on the variables to retain.

Given the considerably small sample size (n = 27) of your dataset, I suggest you consider running Cointegration regression - three options you can choose from based on your sample properties being the following:

(a). Dynamic OLS (DOLS)

(b). Fully-Modified OLS (FMOLS)

(c). Canonical cointegrating regression (CCR).

These methods are designed to best handle small samples, and produce reliable estimates.

ALL the best!

Here is a link

In principle, it is impossible to find out causality only with statistical methods. Statistics can only tell you, if there is a more or less good relation between variables. It is on you to assume causality between variables, this is part of the identification of the model. Luckily, it is mostly quite clear in which direction causality runs (using logics, theories or merely common sense). If you formulate a linear equation for a (assumed) causal relation and you get bad statistical results, you can be sure that the relation is either non-linear or does not exist. That is why I think, that one should think about how the relation (function) may be already in the process of identification.

As, in general, causality needs time (i.e. the "caused" variable reacts with some time lag and the data of a sample are for the same period, one cannot find out the "timing", which might be necessary for testing causality (But this is also true for e.g. annual time series, when the reaction lag is short).

Hello，

In my view, you can use package R "quantreg".

And you can use this command:rq(formula, tau=c(0.05, 0.25, 0.5, 0.75, 0.95))

Hope the website following can help you ! ↓

It depends on how the shock influenced the treated and control unit.

If you can argue that the shock influenced all units in the same way, then the inclusion of two way fixed effects (particularly time fixed effects in this case) can help you to rule out the effects of such 'common' shock.

However, if there are reasons to think that the structural shock differently affected treated and control units, then this could result in potential biases in the estimation and identification of the average treatment effect.

There is a novel literature dealing with the correct identification of such effects. I would suggest you reading the following paper:

"VISUALIZATION, IDENTIFICATION, AND ESTIMATION IN THE LINEAR PANEL EVENT-STUDY DESIGN" from Simon Freyaldenhoven, Christian Hansen, Jorge Pérez Pérez and Jesse M. Shapiro. You can find the paper here: http://www.nber.org/papers/w29170

I particularly suggest you to read section 3.1. of the paper as long as accompanying the paper there is the Stata command "xtevent" which is particularly useful to deal with such econometric analyses.

Hi Comfort Motlhabane , it's soft copy is not available on internet, because its latest publication. Better to search for other alternative or go with paperback.

This study can be helpful in this regard. Good luck and success to all.

The Role of Financial Inclusion in Achieving Stability and Economic Growth.

Look at a scatterplots which may help but it seems that you don't have much of a relationship here. Best wishes David Booth

The difference between R squared in the ARDL model and R squared in the bounds test is that R squared in the ARDL model is about the rubostness of the overall model result and not about the robustness of the bounds test result only. This means R squared in the ARDL model will tell you how much the variability in Y or Ys (your dependent variable/variables) is/are explained by the ARDL model given Xs variables (your independent variables). On the other hand, R squared in the bounds test is specifically about the bounds test result within the ARDL model and not about the robustness of the overall model result. You need an ARDL model first to have a bounds test but not a bounds test to have the ARDL model. I hope it helps you.

If b is 0 for women and 1 for men, then exp(a) is the (geometric) mean wage of women, and men earn on average exp(b)-1 percent more than women. b percent is an underestimation (because exp(b)=1+b+b²/2+b³/6+....), which is negligable for low b's, because the significance intervals of such an estimate may likely be considerably higher. In this example, I would say (write), that men earn about 7% more.

By the way, it is better to accept percent simply as a format of presenting a number, i.e. 0,067 can be shown as 67/1000, 67000 ppm, 6,7/100, 6,7% etc. If you write 0,067 into a cell of EXCEL and choose the format %, it will be shown as 6,7%. One should not use percent=value*100, as we have learned in school, because this could make things more complicated than necessary.

1. to check the endogeneity run a regression and find residuals.

2. then check the correlation between endogenous variable(s) and residuals.

3. if there is a high correlation endogeneity is present otherwise not.

Alternet method usge ivreg and ivendog with Instrumental Variables

I suggest E-views is better to use. I used E-views for panel data in several occasions and I find it easy and better.

If you use E-views, let the lag selection criteria be performed by the function, which means that you are not to select the lag selection criteria, but if you do not get it negative and significant, you decrease your variables. It is always advised to have fewer explanatory variables than more explanatory variables. I hope it helps you.

I have often been approached with questions such as this. I would have known the person and his background in economics, statistics/econometrics, linear algebra, and mathematics. I would also have known or asked why he wanted to proceed in this way and what he hoped to achieve. The lists below are long but I would hope that you find something there that is useful.

I know of three modern introductions to econometrics that are accompanied by R companions.

The first three books cover an undergraduate introduction to econometrics. I found item 5 very useful but you probably need to have a good knowledge of econometric theory to take full advantage of it.

There are also various graduate textbooks on time series analysis, panel data analysis, and various other aspects of econometrics.

Econometrics is often taught as a series of recipes. e.g. estimate the following expression using OLS, without an explanation of the importance of the underlying economic theory. Causality is often inferred from tests of significance without an understanding of how the statistical tests depend on the underlying economic theory. The ideas of tests of hypotheses and p-values are often not understood. The books below are attempts to explain what can be achieved by statistical/econometric analysis and how to avoid false conclusions.

As far system Gmm is concerned, Hansen J is superior to sargan test and it will even give a better result if robust option specified.

I think R software is better and there are many packages (ex. lme4) linear mixed effect models. You can try. I think it is very useful.

You need to take enough data (time period) before and after tax cuts. A good econometric model based on time series or panel data will suffice.

You can have a look at gravity model of trade.

You can use air or water cooling system. Water cooling sytem will be cumbersome and risky. But air cooling system would be safe involving aluminium or copper heat sinks and cooling fans. Much would depend on the amount of heat generated.

Mr a. D.

The ECT(-1)os always the lagged value of your dependent variable.

Regards

Check this out: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwiG0-uH4M34AhVoxYUKHcIoD6EQFnoECBkQAQ&url=https%3A%2F%2Fwww.econstor.eu%2Fbitstream%2F10419%2F73774%2F1%2FIfoWorkingPaper-118.pdf&usg=AOvVaw2s7COMAs8GxtjdDw_TJzIt

Bonjour mon cher,

Je suppose qu'en réalisant votre acp, c'est dans le but construire un score à partir de vos données recueillies au moyen d'une échelle de likert.

Si c'est le cas, je tiens dans un premier temps à vous rassurer que les valeurs obtenues en prédisant le score peuvent être négatives ou positives (comme expliqué plus haut). Cependant, si vous souhaiter avoir des valeurs comprises entre 0 et 1 afin de pouvoir les comparer, vous devez normaliser le score obtenu. L'une des méthodes (il y en a plusieurs) les plus utilisées est celle du min max qui consiste à soustraire le minimum à chaque valeur, puis diviser le tout par l'étendue (max - min). Ainsi, tous vos scores varient entre 0 et 1.

J'ai cependant une réserve sur le choix d'une acp. Les données obtenues à l'aide de l'échelle de likert, bien que codées en chiffres, ne restent pas moins des données catégorielles. C'est a dire qualitatives. De mon point de vue, une amc (analyse en correspondances multiples) serait plus appropriée.

Merci et bonne suite dans vos travaux.

Dear Simon, you confirm my contribution. Statstistics was applied by many sciences long before the term econometrics was invented. Like mathematics (one could, of course, say that statistics is part of mathematics), it is universal in the sense, that it can be applied for rahter different scientific studies. Nevertheless, it can only be applied, if there is a theory of the respective science.

A VAR model is often mixed up with an economic model and used instead of carefully thinking about the specification. That seems to be also the case here. One just takes time series with enough data (I think quarterly) for inflation and unemployment and "throughs" them into a VAR estimation program. That has nothing to do with economics, it is rather the dontrary of good econometrics (neither economics nor sensible statistics).

After putting more brain effort into a good specication (considering other influences, too; see John's advice), you should also look to find out the main lags (see Joel's answer) and whether there are unlagged (in the same quarter) effects, too. You must also take attention that unemployment is a stock (measured at which point of time?), but inflation is a flow.

Standard deviation is a measurement of investment volatility. It measures the performance variation from the average.

Dear Atiqur Rahman, I would like to give you an example from my Polish market and my experience as a business analyst. Well, at the turn of the 20th and 21st centuries in my country, you could see a lot of small and medium-sized companies with a very high rate of growth in revenues and market shares. In many cases in small family businesses, as a result of the investment process and implementation of new technologies, work efficiency increased quickly, but what I found many times when analyzing these "star" companies, such rapid successes in terms of the dynamics of revenue growth, market share and EBIT operating profit caused new challenges and the emergence of internal problems in companies as well as reactions / counter-actions of stronger competition on the part of large international corporations. Well, the best employees who contributed to the increase in the labor productivity index and were directly involved in departments using modern technologies, immediately demanded or expected a salary increase or promotion to a higher position or left the competition. On the other hand, the management of such small family businesses, stating that the company is developing very well because its work productivity increases, immediately tried to use these positive signals to win new orders and to workload its experienced and efficient employees to an even greater extent. Therefore, according to the theory of neoclassical microeconomics, small, fast-growing companies in Poland tried to expand the use of the labor factor, the productivity of which was increasing and overburdened employees with work duties (one of the highest rates in the world of the number of hours worked per month and year per employee according to ILO). This means the macroeconomic environment, that is: high unemployment on the market, low level of social benefits, no possibility (before joining the European Union in 2004) to emigrate to another labor market with better social protection and higher salaries per hour. Entrepreneurs who started using strategies to maximize the use of the labor factor accepted too many new orders and borrowed from banks, taking high-interest loans (high loan costs) and were doing worse and worse in rebuilding the company and adapting to the new conditions of a higher level of production. The news about "squeezing" out of employees as much as possible, while the employer did not raise wages and bonuses, meant that employment in such a company was avoided (employment barrier) and such a company began to lose the most ambitious and experienced people and problems with handling large orders, based on deferred payments in time (typical are construction and assembly works, infrastructure projects). Problems with financial solvency also emerged quickly, as the focus on maximizing the use of the labor factor with high indebtedness in terms of investment loans led to problems with paying the loan installments on time.

Best regards DG

I will add to @Anton Rainer's answer some questions that you should consider.

1 Are there day of the week effects?

2 Are there holiday effects?

3 are the goods perishable or can unsold goods be sold on the following days?

4 Are there any other variables (correlated with an included explanatory variable) that might effect the relationship?

Hie guys...i have N=9 and T=10, which method should I use??

Shahin Behdarvand, there are several books and online resources for undergraduate and graduate students. Please visit https://www.econometricsbooks.com/ for information on these.

Shahin Behdarvand , in my case I simply love MatLab. I have been using R, Python, Eviews, Stata and MatLab for the last 15 years, and I could say that when one wants to estimate somebody else models, R, Python, Eviews, Stata come handy; but in my case when I want to build and estimate my own models (this is, models that I create myself), MatLab is my tool of choice. MatLab is losing popularity in econometrics, and because it is not open as R or Python, most data science laboratories and statistical units that used to apply MatLab are migrating to open tools. Stata keeps being popular despite the fact that you have to pay for use it; Stata has improved a lot true to be said.

I would prefer ind of ANOVA statistical analysis

Hi there, I am using Eventus software and I am wondering how the software computes the Market index in order to calculate abnormal returns?

Can somebody share some papers/literature that state that stationarity is not to be tested when T is less than a certain amount (say 30 or 15, as being discussed above)? It would be extremely helpful for my thesis.

*Panel ARDL

*PMG

xtpmg d.output d.fcapinf d.bankcr d.pfolio d.equity, lr(l.output fcapinf bankcr pfolio equity) ec(ECT) replace pmg

xtpmg d.output d.fcapinf d.bankcr d.pfolio d.equity, lr(l.output fcapinf bankcr pfolio equity) ec(ECT) replace full pmg

*MG

xtpmg d.output d.fcapinf d.bankcr d.pfolio d.equity, lr(l.output fcapinf bankcr pfolio equity) ec(ECT) replace mg

xtpmg d.output d.fcapinf d.bankcr d.pfolio d.equity, lr(l.output fcapinf bankcr pfolio equity) ec(ECT) replace full mg

*Hausman Test to Know Which to Choose (Significant prob. will mean choosing MG and insignificant prob. will mean choosing PMG)

hausman mg pmg, sigmamore

Another way to look at ARDL is that the software can produce nonsense results even after you have done everything correctly. That is, all your variables may be statistically insignificant. It is these diagnostic tests plus most importantly, your eyes that will show you that something else is needed. Always run ARDL as an OLS model before accepting what software generates.

For a 1 unit increase in Health expenditure, the Infant mortality rate decreases by -0.39.

You can yes. You can also use Eviews or Stata softwares

Maria Rabbi ,

I would say:

Regarding the interpretation. Let's assume that b0 is the coefficient of "Years to maturity" and b1 is the coefficient of "Years to maturity x COVID".

b0 is the effect of years to maturity in the pre-covid period.

b0 + b1 is the effect of years to maturity in the post-covid period.

b1 captures the differential effect of years to maturity in the second period compared to the first period.

I hope it helps,

José-Ignacio

My specific area is Panel Data.

The software I am using Is EVIEWS9.

I am doing both I will give guidance to learners as well as sometimes i ask for guidance

Gopal Bihari Saraswat

I Hope this book will help you

You might want to check this chapter:

The estimated model AR(1)-GARCH(1,1) is different from the true model AR(3)-GARCH(1,1). The estimates must therefore be biased.

Daniel Wright and Remal Al-Gounmeein thanks for the links, I will take a look. We are matching respondents based on their likert scale (5) responses to 16 partisan political positions.

recent development

(1) Wooldridge two-way Mundlak regression and fixed effects and dif-in-dif

(2) synthetic control

(3) Cerulli, G. 2015. Econometric Evaluation of Socio-Economic Programs: Theory and Applications.

(4) Pesaran (2015) Time Series and Panel Data Econometrics

In fact, please note that your topics are mutuall inter-related and inter-connected. These are all based on the fundamental theme of sustainability. The concept of 'Smart' / 'Smartness' arises because of the maximum use of renewable sources of energy, minimising the carbon footprint, ensuring minimum harm to the environment, strict adherence to Green Building norms, effective and efficient use of ICT, and so on. These ensure the fundamental concept of Sustainable Development (SD). In this regard, you may just go through the SDGs and MDGs too, please.

With regards

Try PhD in Bioinformatics

Dear, conducting any standard model depends on an important factor, namely the number of observations included in the model, for example, if the observations are small, the Phelps-Peron test can be conducted to test the stability, and if the observations are large, the whooping-full test can be conducted, and in light of the stability results, we can determine the model that can be conducted Julius Hogan

Before deciding on the econometric model, you should go through the stationarity test (ADF test). If the data are stationary, OLS Regression with a log-linear model would be fine. But, if not, you may go for VAR or ARDL. You should also check the robustness of the model by going through residual tests such as Autocorrelation LM Test.

It seems a very small sample to apply microeconometric techniques. Having 27 seven observations and 10 covariates, at most, you will have 27 - 1 - 10 = 16 degrees of freedom. This is pretty low. If I had to decide to pursue a project based on that, I would try to avoid it.

It is really closer to multiple times series than panel data. Have a look at this link:

Just perform a regression of the variable on time or a simple correlation.

Neverthless, usually, what we do is to carry out a test for mean differences at two different points of time.

I notice the following about your specification. (1) Your inclusion of a constant (and its subsequent significance) means you allow for (and find) a trend in the real exchange rate independent of any trends in the other variables. Is that economically reasonable? (2) I assume the CRISIS variable is a zero-one dummy for time periods with a "crisis" of some sort. Apparently it is not in the cointegration vector. Why not? If it were, then I'd expect to find CRISIS differences in the error correction equation. Instead, you have it in levels. Thus you specify that a temporary crisis has a permanent effect of the level of the real exchange rate independent of the other variables. Is that what you intend? (3) You do not include the lagged difference of the real exchange rate in the error correction equation. Why not? Normally it would be there.

I may be misunderstanding your question, but there is no constant variance assumption with logistic regression, so you do not need to worry about heteroskedasticity. In fact, heteroskedasticity is almost guaranteed with logistic regression since the variance of a binomial random variable is a function of the probability of the event happening and the probability of the event not happening, which will usually differ between observations.

Dear Gonçalo Maria ,

Please note that transforming variable(s) does NOT make the series stationary, but rather makes the distribution(s) symmetrical. Application of logarithmic transformation needs to be exercised with extreme caution regarding properties of the series, underlying theory and the implied logical/correct interpretation of the relationships between the dep variable and associated selected regressors.

Reverting to your question, the proposed solution would be to use the Autoregressive Distributed Lag (ARDL) model approach, which is suitable for datasets containing a mixture of variables with different orders of integration. Kindly read the manuscripts attached for your info.

All the best!

Hello Robert Traistaru ,

It is shown as lnpop, so i think you need to transform the variables into log. Simply follow this steps if it is the case,

quick<generate series<lnpop= log(pop). and log the rest of the variables in the same way.

after that put lnpop in the estimate equation.

Best wishes for you.

Ibrahim Niftiyev , probably the reviewer wants to see not only whether a variable affects or not the other (i.e. the results of the Granger causality tests), but also to which extent (the magnitude and temporality of the dynamic relationship, something you can obtain from the IRFs of a VAR model). If you want to apply a VAR but you have a small sample size/data limitations, you want to consider a Bayesian VAR. Bayesian VARs are very popular and Bayesian methods are valid in small samples.

The Ljung-Box test is aimed at testing the independance of errors using residuals of an ARMA model estimated on the same data. But it makes use of autocorrelations so it is not powerful when the errors are uncorrelated but not independent. When applied to squared residuals, it can reveal ARCH and GARCH effects. Note that the errors of a ARCH-GARCH model are uncorrelated but not independent. Have a look at the excellent book by Francq and Zakoian entitled "GARCH Models: Structure, Statistical Inference and Financial Applications" published by Wiley in 2010.

You may find this attached paper helpful.

Best wishes, David Booth

Robert Traistaru , GLS was proposed as a solution to the presence of autocorrelation or heteroscedasticity, which violates the OLS assumption that the error terms are uncorrelated and with constant variance, meaning that the Gauss Markov theorem does not apply and hence the OLS estimators are no longer BLUE, i.e. the estimates are inefficient. In short, the justification for using GLS is the presence of some form of autocorrelation or heteroscedasticity in the errors.

However, when you do not know the value of rho (needed for GLS), and you need to estimate the value (which tends to be the case), then you apply in practice Estimated GLS (EGLS), aka FGLS. While GLS is more efficient than OLS under heteroscedasticity or autocorrelation, EGLS is only asymptotically more efficient, but in small or medium sample sizes, it can be less efficient than OLS. So, if you have a moderate to small sample size, using EGLS is actually worst than using OLS.

I am also interested to the answers of this topic.