About ets_kps

ets_kps · ‎11-20-2013

I have a couple of suggestions. 1) Post the data and the example SAS file you are working on. (so we know where this is falling down) 2) Repost this in Forecasting and Econometrics as the "right" eyeballs are likely to see it over there. Best-Ken

ets_kps · ‎11-20-2013

This is likely because Jan is not in the model as it was dropped due to the "dummy variable trap." Make sure your tests refer to parameters that were actually estimated. -Ken

ets_kps · ‎11-08-2013

Did you come across these examples? Methods Matter: Improving causal Inference in Educational and Social Science Research

ets_kps · ‎11-05-2013

Hi Allie, In addition to Udo's request I thought i'd push you in a certain direction. The types of multiple seasonal/cycle models can be estimated in SAS using a UCM framework. You can think of this as ESM-like but with regressors. These regressors might be continous, such as a price, or dichotomous, such as a day. You might want to experiment a bit with the combination of seasonal length as well as some dummies for month to take care of those cycles. The dummy variable approach will require a good deal of data but as we don't know what you are working with, I'll just toss some ideas your way. SAS/ETS(R) 12.1 User's Guide Udo, might have more to add with some data. Regards-Ken

ets_kps · ‎10-31-2013

There are many directions to take your question but one technique you might want to consider is Stochastic Frontier Modeling. The underlying model assumes there is a maximum efficiency that can be obtained. Agents in the model can be evaluated based on their "closeness" to maximum efficiency. Here is a link to its implementation in SAS. SAS/ETS(R) 12.3 User's Guide Also, here a link to paper you might mind useful. http://isbm.smeal.psu.edu/library/working-paper-articles/2004-working-papers/06-2004-measuring-performance.pdf It doesn't sound as if your data are confidential. Would you like to share them?

ets_kps · ‎10-28-2013

Hello Amlan, Thank you for your question. Actually, I took your file and code and tried it on my 9.4 SAS with 12.3 SAS/ETS and the optimization completed just fine. My guess is that you are on a previous version of SAS/ETS. In recent years, the VARMAX routine has received algorithm improvements that greatly enhance optimization. If you are unable to upgrade to the newest version of SAS you might want to see if you can bump up the iterations and function calls, such as with nloptions maxiter=500 maxfunc=5000; Best of luck-Ken

ets_kps · ‎10-25-2013

Hi Stephane, Yes, these output differences occur from time to time. Sorry it has been a trouble spot for you. To your second question, could you clarify? Are you looking to have SAS/ETS 12.1 documentation be the default home page? Rather than 9.x? thanks

ets_kps · ‎09-27-2013

here is a response from the developer: Given one data set containing two columns, exchangeRate and stockReturn (will need by groups for all countries in your example): Proc varmax data=one; Model exchangeRate stockReturn = / p=1; /* VAR(1) with constant mean */ Garch p=1 q=1 form=BEKK; /* BEKK GARCH(1,1) */ Run; Note that in proc varmax, we calculate constant term, C’C, in garch equation in equation (2) as one symmetric matrix. If the user wants to repeat the paper, he can apply Cholesky decomposition on the constant matrix (by using IML for example). I hope this helps. Also, please contact SAS tech support should you require additional help getting started with VARMAX.

ets_kps · ‎09-12-2013

Well, the first thing you need to do is reshape your data from WIDE to LONG with PROC TRANSPOSE. proc transpose data=wide out=long; by date; run; (the first post doesn't need this) Make sure your date variable is formatted as a date and then you can use the following code to accomplish the 12 month percentage change. It uses a new ETS procedure called TIMEDATA (h/t to the TIMEDATA developers for this). The other operations will be very similar. Just a little background on TIMEDATA... It is extremely efficient at the types of time series operations that you are trying to do. data verylarge; set sashelp.air; clientid=1; balances=air; run; proc timedata data=VERYLARGE out=_NULL_ outarray=trend plot=arrays; by clientid; id date interval=month accumulate=average; var balances; outarray trend; do t=1 to _LENGTH_; if t < _SEASONALITY_ then trend = .; else do; previous = balances[t-_SEASONALITY_]; recent = balances ; if previous = 0 then trend = .; else trend = (recent - previous) / previous; end; end; run;

ets_kps · ‎08-19-2013

Hi Mayank, Thanks for the question. Would you mind sharing some code for what you are doing? Perhaps a graph of the series or some background on the type of units? There are a number of simple examples at SAS/ETS(R) 12.1 User's Guide which could help to get you started. Thanks

ets_kps · ‎07-19-2013

Hi Niam, Thanks for the question. Yes, SAS has a number of procedures for nonlinear panel data models. I am going to TCOUNTREG: Count data models (Poisson and Negative Binomial) with Fixed and Random Effects (subject specific intercepts)SAS/ETS(R) 12.1 User's Guide NLMIXED: Syntax and models similar to PROC MIXED but non-linear models. SAS/STAT(R) 12.1 User's Guide You can also analyze repeated measures with GLIMMIX. I will warn you. Since you used the term "panel" and not "longitudinal" to describe the data, I am going to assume you are coming from more of an observational data background. For that reason some of the terminology of the *MIX*-type procedures may look a little odd at first glance. Let me know if you need any assistance getting started-Ken

ets_kps · ‎04-11-2013

In addition to Udo's reponse I would also ask what your particular objective is? If it is prediction, then perhaps not finding a huge response is acceptable. If, on the other hand, your objective is to find the marginal effect of the program then you might either re-examine the tool (PROC AUTOREG) or the functional form (lags: the effect might not be contemporaneous). My two cents.

ets_kps · ‎04-11-2013

Hi Jing, I spoke with the developer of proc model and he had this response, " For the SUR method MODEL computes a pseudo inverse of the OLS residuals’ covariance matrix. When the OLS covariance matrix is singular a warning is produced explaining which equations’ rows in the covariance matrix are linearly dependent, and PROC MODEL continues with the SUR estimation of parameters. If the user isn’t getting this warning message for a known deficient matrix then something else is going on " He also requested that he would be happy to look at your code and data if you wouldn't mind providing and example of the issue. - Ken

ets_kps · ‎03-25-2013

There is an alternative way to do this that doesn't involve using all this first., last. syntax. Using PROC TIMEDATA, you can directly reference the time series within BY groups. I feel the syntax is much more intuitive. Here is full sample in action. data greene; input firm year production cost @@; date = mdy(1,1,year); datalines; 1 1955 5.36598 1.14867 1 1960 6.03787 1.45185 1 1965 6.37673 1.52257 1 1970 6.93245 1.76627 2 1955 6.54535 1.35041 2 1960 6.69827 1.71109 2 1965 7.40245 2.09519 2 1970 7.82644 2.39480 3 1955 8.07153 2.94628 3 1960 8.47679 3.25967 3 1965 8.66923 3.47952 3 1970 9.13508 3.71795 4 1955 8.64259 3.56187 4 1960 8.93748 3.93400 4 1965 9.23073 4.11161 4 1970 9.52530 4.35523 5 1955 8.69951 3.50116 5 1960 9.01457 3.68998 5 1965 9.04594 3.76410 5 1970 9.21074 4.05573 6 1955 9.37552 4.29114 6 1960 9.65188 4.59356 6 1965 10.21163 4.93361 6 1970 10.34039 5.25520 ; proc sort data=greene out=greene; by firm year; run; proc timedata data=greene out=differenced; by firm; id date interval=year ; var cost /dif=(5); var production /diff=5; run; Yours might look like; PROC TIMEDATA data=returns out=returns2; by company; id Date interval=day; var price /diff=(1); run; There are also ways to do this in PROC PANEL

ets_kps · ‎03-20-2013

Just to add a bit to Udo's response.... There are full, runnable examples of UCM and SSM here. This example seems closest to your needs. Best of luck-Ken

Online Status	Offline
Date Last Visited	‎10-09-2015 06:19 PM

Re: Interval option in MARKETDATA (RiskDimension)

Re: proc panel and heteroscedasticity correction

Re: proc panel and heteroscedasticity correction

Re: PROC IMPORT from Web

PROC IMPORT from Web

Re: 2-stage least squares regression

Re: poisson regression using sas

Re: poisson regression using sas

Re: poisson regression using sas

Re: poisson regression using sas

Re: Implying white's (1980) correction at proc panel procedure

Re: poisson regression using sas

Re: Proc Similarity to cluster time series data

Re: 2-stage regression with more than 2 models

Re: Proc Panel and Parks

Re: Proc ESM Seasonal Model Help

Okay ODS, just give me what I want!

Hausman-Taylor estimator for SAS

Re: please, help me with GMM estimation on asset pricing model

Re: partial F-test in PROC PANEL

Re: How to use Sharp Regression Discontinuity (SRD) in SAS?

Re: Proc ESM Seasonal Model Help

Re: Building an index of performance

Re: VAR(1) ARCH(1) model

Re: Proc Panel fixone and fixtwo

Re: GARCH BEKK with example

Re: Trend Analysis- SAS Logic?

Re: UCM modelling

Re: Nonlinear regression models with panel data

Re: PROC UCM estimates

Re: SUR (seemingly unrelated regression) in SAS

Re: how to compute returns if the panel data has continuous observatio...

Re: Proc Statespace Procedure