Robust Feasible Generalized Least Squares FGLS Model in STATA

แชร์
ฝัง
  • เผยแพร่เมื่อ 10 พ.ย. 2024

ความคิดเห็น • 59

  • @irfanrashidganie3832
    @irfanrashidganie3832 2 ปีที่แล้ว +2

    When I run xtgls depvar indepvar...., panels(corr) corr(ar1) I get this error "year is not regularly spaced or does not have intervals of delta -- use the force option to treat the intervals as though they were regular
    "
    then I use force at the end but still shows this error "panels must be balanced". I have already used "xtset code year" to define it is a panel data.

    • @nomanarshed
      @nomanarshed  2 ปีที่แล้ว +1

      th-cam.com/video/hMuMst3k1YQ/w-d-xo.html this video will help in filling gaps

  • @waititstiiina
    @waititstiiina 5 หลายเดือนก่อน

    Hello , After running the fgls model how can I do post estimation diagnostic tests? To check of the problems have been solved ?

    • @nomanarshed
      @nomanarshed  5 หลายเดือนก่อน

      There are no direct post estimation diagnostics available. what people have done is they generate the residuals and check manually like sktest for normality, AR1 equation based DW test for autocorrelation, BG test using squared residuals as dv for heteroskedasticity and adding estimated dependent squared in the model to check for RESET test. other tests include cross sectional dependence on residuals. etc

  • @pavilionchannel9492
    @pavilionchannel9492 6 หลายเดือนก่อน

    Hi Dr. Which result should we interpret after remove all problems? Is it the last output?

    • @nomanarshed
      @nomanarshed  6 หลายเดือนก่อน

      Yes . The output which has no problems

    • @pavilionchannel9492
      @pavilionchannel9492 6 หลายเดือนก่อน

      Thanks Dr.

  • @hossainaisjust
    @hossainaisjust 8 หลายเดือนก่อน

    Helpful video. My data is unbalanced panel and panels(corr) is not working. But there is cross-sectional correlation. What can I do?

    • @nomanarshed
      @nomanarshed  8 หลายเดือนก่อน +1

      The panels(corr) is actually for timeseries correlation and it is applied on balanced data only. Try estimating driscoll and kraay regression which is robust to crosssectional regression.

  • @kalemguy
    @kalemguy ปีที่แล้ว

    How about to look for the R square on this FGLS ( xtgls depvar indepvar...., panels(corr) corr(ar1))? If there isn't what is to use for the replacement?

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      In advanced methods, R squared is not a good criteria to judge the value of a model. Any ways, you can manually generate it by generating residuals.

  • @mohsinsmir3104
    @mohsinsmir3104 ปีที่แล้ว

    Wat to do if we only have heteroscadetisity and auto correlation but no cross sectional dependence.

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      This FGLS model is suited for that. It does not work when cross sectional dependence is present

  • @MackintoshWright
    @MackintoshWright หลายเดือนก่อน

    5339 Barton Extension

  • @TrịnhThụyÁiTrâm
    @TrịnhThụyÁiTrâm 8 หลายเดือนก่อน

    Hi, I used the code crest but it said “too many variables specified”. How can I deal with it? Thank you so much

    • @nomanarshed
      @nomanarshed  8 หลายเดือนก่อน

      Can you share the code here

  • @sarahahmedchawsheen5455
    @sarahahmedchawsheen5455 ปีที่แล้ว

    is fgls model requires cointegration an unit root test, and what are the postestimation tests after running fgls model? thanks.

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว +1

      This is not a dynamic model so it does not require unit root and countegration test. In post estimation hetroscedasticity, autocorrelation and cross sectional dependence tests needed to be checked

    • @sarahahmedchawsheen5455
      @sarahahmedchawsheen5455 ปีที่แล้ว

      @@nomanarshed thanks Dr. Noman, I appreciate your reply .

  • @FarhatAzad-ix8ze
    @FarhatAzad-ix8ze 6 หลายเดือนก่อน

    Sir when I am running FGLS command with panels(correlated ) in stata 17 I am getting an error r(459) panels must be balanced
    my panel is already strongly balanced
    plz if you can help me

    • @nomanarshed
      @nomanarshed  6 หลายเดือนก่อน

      if stata says data is strongly balanced it only means that each cross section has equal number of years, but for data to be balanced you need to make sure there are no missing values in the data. some values can still be dropped if you had taken log to negative values.

  • @Dr_Shiny
    @Dr_Shiny ปีที่แล้ว

    can we use xtgls with FE, i.e. i.year??
    E.g.
    xtgls depvar Indeptvars, fe

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว +1

      i.id for cross section FE and i.year for time FE can be used in xtgls.

  • @TheresaRobin-c5d
    @TheresaRobin-c5d หลายเดือนก่อน

    584 Durgan Meadows

  • @ananyajaiswal2263
    @ananyajaiswal2263 ปีที่แล้ว

    Hello sir, how to correct heteroskedasticity and cross sectional dependence for unbalanced panel data? When I run xtgls depvar indepvar...., panels(correlated) corr(independent). I am getting error that "panels must be balanced". In my case N is 485 and T is 21. My data is highly unbalanced. It is not possible to interpolate. What to do?

    • @ananyajaiswal2263
      @ananyajaiswal2263 ปีที่แล้ว

      Please help me with problem

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      XTGLS only helps in time series autocorrelation not the cross sectional autocorrelation (cross sectional dependence). In your case you should use Panel Corrected Standard Error Regression or Driscoll and Kraay regression.

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      shared

  • @RenukaKumawat-c9p
    @RenukaKumawat-c9p ปีที่แล้ว

    @noman Arshed can we use FGLS... in case when no. of cross-sections are more than no. of time dimensions (N>T) ?

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      It is preferable when T>N but studies have used this model for N>T

    • @RenukaKumawat-c9p
      @RenukaKumawat-c9p ปีที่แล้ว

      @@nomanarshed Thanks for replying, then how we can justify in this case, coz both cases (N>T and T>N) are there in literature?

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      If it is a paper just add references. If it is a phd thesis you can do one more type of estimation for robutness analysis.

    • @RenukaKumawat-c9p
      @RenukaKumawat-c9p ปีที่แล้ว

      @@nomanarshed Thanks, it definitely helps.

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      Welcome

  • @momk5
    @momk5 2 ปีที่แล้ว +1

    Interesting
    Thanks !

  • @rosananicoleaguirre6727
    @rosananicoleaguirre6727 ปีที่แล้ว

    Can I use FGLS with Fixed effects at the same time?

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      Yes using LSDV method add dummies of your cross sections. It will make your model fixed effect.

  • @Barakat7elfarra
    @Barakat7elfarra ปีที่แล้ว

    Please if there is no cross sectional dependence, we still can use FGLS, and is ok to use it also while there is multicollinearity problem between two variable in the model?

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว +1

      FGLS model is only applicable for stationary data with no dependence. While all models are estimateable under moderate multicollinearity. It will make the estimates biased. One should check if multicollineairty is ignorable.

    • @Barakat7elfarra
      @Barakat7elfarra ปีที่แล้ว

      @@nomanarshed Thanks very much drNoman Arshed, now my data the dependent variable has no cross-sectional dependence, and my independent variable has cross-sectional dependence. in this case what can i do with FGLS model, Sorry to bother you.
      Jazzak Allah Dr..

  • @AnthonyJohnson-m4d
    @AnthonyJohnson-m4d หลายเดือนก่อน

    088 Nicolette Row

  • @raufkhan5846
    @raufkhan5846 10 หลายเดือนก่อน

    hi sir i believe ill get a reply, My study is related to only one country and I have monthly data for 9 variables where 1 variable is dependent rest are independent please suggest me the methodology how to know the relationship for this time series data

    • @nomanarshed
      @nomanarshed  10 หลายเดือนก่อน

      how many total months are there. You can explore this video for symptomatic selection of time series model th-cam.com/video/BspVpx0oHHY/w-d-xo.html

  • @alikaraca163
    @alikaraca163 ปีที่แล้ว

    What if there is only cross-sectional correlation (i.e., cross-sectional dependency) in the model without autocorrelation and heteroscedasticity, how can we fix that problem in this case?

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว +1

      Significant cross sectional correlation is cross sectional dependence. In that case you should you driscoll kray model.

    • @alikaraca163
      @alikaraca163 ปีที่แล้ว

      @@nomanarshed what about just applying " xtgls y1 x1 x2" command? Don't you think it does solve cross-sectional dependency automatically in panel data?

    • @alikaraca163
      @alikaraca163 ปีที่แล้ว

      @@nomanarshed "xtscc P1 MktRF SMB HML Mom D2 D3 D4 D5 D6" P1 is my dependent variable and Ds represent dummy variables. Can I use it that way then? Thank you

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว +1

      Its documentation does not make it robust to any thing it this command. You need to add robustness after comma

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว +1

      Yes. This is the way to make robust for cs dependence

  • @vuquynhanh4345
    @vuquynhanh4345 ปีที่แล้ว

    I get this error "year is not regularly spaced or does not have intervals of delta -- use the force option to treat the intervals as though they were regular". Time variable is from 2017 to 2021 and it has no gaps. I do not find any problem with column "year". Before that, I have used xtset id year to define panel data. Can you help me with this problem?

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      Try to use the solution in this video otherwise your year variables do not have common pattern, then in this case you cannot use time series regression, you can use time dummies in simple cross sectional regression. th-cam.com/video/H6hjYMxEAvU/w-d-xo.html

    • @vuquynhanh4345
      @vuquynhanh4345 ปีที่แล้ว +1

      @@nomanarshed Thank you, I'm so grateful for that

  • @HarrimanSpencer
    @HarrimanSpencer หลายเดือนก่อน

    241 Sammie Ports

  • @Ganieirfan
    @Ganieirfan ปีที่แล้ว

    is GMM better than FGLS? Whaat are the conditions where we can use GMM over FGLS?

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      Cannot compare GMM is used to solve endogeneity and GMM is sensitive to heteroskedasticity. FGLS is used to make model robust to heteroskedasticity. Both models can be used in panel data.

    • @Ganieirfan
      @Ganieirfan ปีที่แล้ว

      @@nomanarshed Thanks for the reply sir. Can you please let me know what are the conditions to use FGLS over fe or re ? Can we also use industry dummy in FGLS?

    • @nomanarshed
      @nomanarshed  ปีที่แล้ว

      @@Ganieirfan FGLS is used when N > T and the data may have heteroskedasiticty and autocorrelation. FGLS has better ability to be robust against FE RE

    • @ivansteven7631
      @ivansteven7631 9 หลายเดือนก่อน

      @@nomanarshed Thank you for your informative videos. I believe for the FGLS model, T > N. I have a close similar question about choosing between the FGLS and ARDL models, given my data's characteristics (T > N, heteroskedasticity).
      Could you advise on criteria for selection, considering research objectives versus elimination criteria? Additionally, can the FGLS model be used when some variables are stationary at I(1) and I(0) and there is cointegration between them?
      I'm sorry if my question sounds basic to you. I am new to these types of models.
      I appreciate your insight.