Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • "can't find uphill direction" when using garch loop working on panel data

    I am using a panel data of n=478 and t=1988-2014, daily data. I use GARCH model and loop code. Each time when running the regression, the following appears, can somebody please help me figure out?

    my code is:
    Code:
    forvalues i=1 (1) 478 {
    arch ut rmrf smb hml umd if id==i', earch(1) egarch(1) distribution(t)
    predict variance if id==i', variance
    replace var2=variance if id==`i'
    drop variance
    }
    every time, the loop are stopped because of:
    flat log likelihood encountered, cannot find uphill direction
    I tried different distribution (normal, t, ged), different garch model, like GARCH(1,1), EGARCH(1,1), OR EGARCH(1,2), all of them cannot work through all panel data.

    P.S. I used code to drop missing data before doing the loop garch

    I really grateful if someone could help me to address this problem.

    Many thanks!!!

  • #2
    Your code has errors in it regarding sections with id==i', rather than id==`i'. Also, you are running a model for each panel. Is that intended?

    Comment


    • #3
      sorry, it’s my type mistake. I doubt checked my do.file, i write both ‘ there. Yes, I ask stata to run GARCH model for each id(portfolio)

      Comment


      • #4
        I have no experience with GARCH models. However, consider this. Each panel has maybe 26 observations. You have 4 independent variables. I bet there is at least one additional parameter in your GARCH model. In general, a very flat log likelihood can be one symptom that a model is unidentified - there is not enough data in the model to estimate all the parameters. Again, are you sure you should be fitting one GARCH model to each portfolio? Because I suspect there is not enough data to do that.
        Please use the code delimiters to show code and results - use the # button on the formatting toolbar, between the " (double quote) and <> buttons.

        Please use the command -dataex- to show a representative sample of data; it is installed already if you have Stata 14.2 or 15.1, else you can install it by typing

        Code:
        ssc install dataex

        Comment


        • #5
          Dear Weiwen,

          I am really appreciate for your reply.

          actually, I use daily data over 26 years. I know GARCH require a large observations to regress, so I even try to use daily data rather than monthly.

          (setting optimization to BHHH)
          Iteration 0: log likelihood = 13241.969
          Iteration 1: log likelihood = 17580.051
          Iteration 2: log likelihood = 17828.4
          Iteration 3: log likelihood = 17929.341
          Iteration 4: log likelihood = 18214.033
          (switching optimization to BFGS)
          Iteration 5: log likelihood = 18304.953
          Iteration 6: log likelihood = 18321.197
          Iteration 7: log likelihood = 18321.681
          Iteration 8: log likelihood = 18323.633
          Iteration 9: log likelihood = 18324.354
          Iteration 10: log likelihood = 18324.7
          Iteration 11: log likelihood = 18324.719
          Iteration 12: log likelihood = 18324.721
          Iteration 13: log likelihood = 18324.722
          Iteration 14: log likelihood = 18324.722
          (switching optimization to BHHH)
          Iteration 15: log likelihood = 18324.722
          Iteration 16: log likelihood = 18324.722 (backed up)
          Iteration 17: log likelihood = 18324.722 (backed up)
          Iteration 18: log likelihood = 18324.722 (backed up)
          Iteration 19: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          Iteration 20: log likelihood = 18324.722 (backed up)
          Iteration 21: log likelihood = 18324.722 (backed up)
          Iteration 22: log likelihood = 18324.722 (backed up)
          Iteration 23: log likelihood = 18324.722 (backed up)
          Iteration 24: log likelihood = 18324.722 (backed up)
          Iteration 25: log likelihood = 18324.722 (backed up)
          Iteration 26: log likelihood = 18324.722 (backed up)
          Iteration 27: log likelihood = 18324.722 (backed up)
          Iteration 28: log likelihood = 18324.722
          Iteration 29: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 30: log likelihood = 18324.722 (backed up)
          Iteration 31: log likelihood = 18324.722 (backed up)
          Iteration 32: log likelihood = 18324.722 (backed up)
          Iteration 33: log likelihood = 18324.722 (backed up)
          Iteration 34: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (0)
          Iteration 35: log likelihood = 18324.722 (backed up)
          Iteration 36: log likelihood = 18324.722 (backed up)
          Iteration 37: log likelihood = 18324.722 (backed up)
          Iteration 38: log likelihood = 18324.722 (backed up)
          Iteration 39: log likelihood = 18324.722 (backed up)
          Iteration 40: log likelihood = 18324.722 (backed up)
          Iteration 41: log likelihood = 18324.722 (backed up)
          Iteration 42: log likelihood = 18324.722 (backed up)
          Iteration 43: log likelihood = 18324.722 (backed up)
          Iteration 44: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 45: log likelihood = 18324.722 (backed up)
          Iteration 46: log likelihood = 18324.722
          Iteration 47: log likelihood = 18324.722 (backed up)
          Iteration 48: log likelihood = 18324.722 (backed up)
          Iteration 49: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          Iteration 50: log likelihood = 18324.722 (backed up)
          Iteration 51: log likelihood = 18324.722
          Iteration 52: log likelihood = 18324.722 (backed up)
          Iteration 53: log likelihood = 18324.722 (backed up)
          Iteration 54: log likelihood = 18324.722
          Iteration 55: log likelihood = 18324.722 (backed up)
          Iteration 56: log likelihood = 18324.722 (backed up)
          Iteration 57: log likelihood = 18324.722
          Iteration 58: log likelihood = 18324.722 (backed up)
          Iteration 59: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 60: log likelihood = 18324.722 (backed up)
          Iteration 61: log likelihood = 18324.722 (backed up)
          Iteration 62: log likelihood = 18324.722 (backed up)
          Iteration 63: log likelihood = 18324.722 (backed up)
          Iteration 64: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (1)
          Iteration 65: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (2)
          Iteration 66: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (3)
          Iteration 67: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (4)
          Iteration 68: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (5)
          Iteration 69: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (6)
          Iteration 70: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (7)
          Iteration 71: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (8)
          Iteration 72: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (9)
          Iteration 73: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (10)
          Iteration 74: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 75: log likelihood = 18324.722 (backed up)
          Iteration 76: log likelihood = 18324.722 (backed up)
          Iteration 77: log likelihood = 18324.722 (backed up)
          Iteration 78: log likelihood = 18324.722 (backed up)
          Iteration 79: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (11)
          Iteration 80: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (12)
          Iteration 81: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (13)
          Iteration 82: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (14)
          Iteration 83: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (15)
          Iteration 84: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (16)
          Iteration 85: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (17)
          Iteration 86: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (18)
          Iteration 87: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (19)
          Iteration 88: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (20)
          Iteration 89: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 90: log likelihood = 18324.722 (backed up)
          Iteration 91: log likelihood = 18324.722 (backed up)
          Iteration 92: log likelihood = 18324.722 (backed up)
          Iteration 93: log likelihood = 18324.722 (backed up)
          Iteration 94: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (21)
          Iteration 95: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (22)
          Iteration 96: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (23)
          Iteration 97: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (24)
          Iteration 98: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (25)
          Iteration 99: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (26)
          Iteration 100: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (27)
          Iteration 101: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (28)
          Iteration 102: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (29)
          Iteration 103: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (30)
          Iteration 104: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 105: log likelihood = 18324.722 (backed up)
          Iteration 106: log likelihood = 18324.722 (backed up)
          Iteration 107: log likelihood = 18324.722 (backed up)
          Iteration 108: log likelihood = 18324.722 (backed up)
          Iteration 109: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (31)
          Iteration 110: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (32)
          Iteration 111: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (33)
          Iteration 112: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (34)
          Iteration 113: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (35)
          Iteration 114: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (36)
          Iteration 115: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (37)
          Iteration 116: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (38)
          Iteration 117: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (39)
          Iteration 118: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (40)
          Iteration 119: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 120: log likelihood = 18324.722 (backed up)
          Iteration 121: log likelihood = 18324.722 (backed up)
          Iteration 122: log likelihood = 18324.722 (backed up)
          Iteration 123: log likelihood = 18324.722 (backed up)
          Iteration 124: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (41)
          Iteration 125: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (42)
          Iteration 126: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (43)
          Iteration 127: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (44)
          Iteration 128: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (45)
          Iteration 129: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (46)
          Iteration 130: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (47)
          Iteration 131: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (48)
          Iteration 132: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (49)
          Iteration 133: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (50)
          Iteration 134: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 135: log likelihood = 18324.722 (backed up)
          Iteration 136: log likelihood = 18324.722 (backed up)
          Iteration 137: log likelihood = 18324.722 (backed up)
          Iteration 138: log likelihood = 18324.722 (backed up)
          Iteration 139: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (51)
          Iteration 140: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (52)
          Iteration 141: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (53)
          Iteration 142: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (54)
          Iteration 143: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (55)
          Iteration 144: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (56)
          Iteration 145: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (57)
          Iteration 146: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (58)
          Iteration 147: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (59)
          Iteration 148: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (60)
          Iteration 149: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 150: log likelihood = 18324.722 (backed up)
          Iteration 151: log likelihood = 18324.722 (backed up)
          Iteration 152: log likelihood = 18324.722 (backed up)
          Iteration 153: log likelihood = 18324.722 (backed up)
          Iteration 154: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (61)
          Iteration 155: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (62)
          Iteration 156: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (63)
          Iteration 157: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (64)
          Iteration 158: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (65)
          Iteration 159: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (66)
          Iteration 160: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (67)
          Iteration 161: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (68)
          Iteration 162: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (69)
          Iteration 163: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (70)
          Iteration 164: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 165: log likelihood = 18324.722 (backed up)
          Iteration 166: log likelihood = 18324.722 (backed up)
          Iteration 167: log likelihood = 18324.722 (backed up)
          Iteration 168: log likelihood = 18324.722 (backed up)
          Iteration 169: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (71)
          Iteration 170: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (72)
          Iteration 171: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (73)
          Iteration 172: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (74)
          Iteration 173: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (75)
          Iteration 174: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (76)
          Iteration 175: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (77)
          Iteration 176: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (78)
          Iteration 177: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (79)
          Iteration 178: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (80)
          Iteration 179: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 180: log likelihood = 18324.722 (backed up)
          Iteration 181: log likelihood = 18324.722 (backed up)
          Iteration 182: log likelihood = 18324.722 (backed up)
          Iteration 183: log likelihood = 18324.722 (backed up)
          Iteration 184: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (81)
          Iteration 185: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (82)
          Iteration 186: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (83)
          Iteration 187: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (84)
          Iteration 188: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (85)
          Iteration 189: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (86)
          Iteration 190: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (87)
          Iteration 191: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (88)
          Iteration 192: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (89)
          Iteration 193: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (90)
          Iteration 194: log likelihood = 18324.722 (backed up)
          (switching optimization to BHHH)
          Iteration 195: log likelihood = 18324.722 (backed up)
          Iteration 196: log likelihood = 18324.722 (backed up)
          Iteration 197: log likelihood = 18324.722 (backed up)
          Iteration 198: log likelihood = 18324.722 (backed up)
          Iteration 199: log likelihood = 18324.722 (backed up)
          (switching optimization to BFGS)
          BFGS stepping has contracted, resetting BFGS Hessian (91)
          Iteration 200: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (92)
          Iteration 201: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (93)
          Iteration 202: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (94)
          Iteration 203: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (95)
          Iteration 204: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (96)
          Iteration 205: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (97)
          Iteration 206: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (98)
          Iteration 207: log likelihood = 18324.722 (backed up)
          BFGS stepping has contracted, resetting BFGS Hessian (99)
          Iteration 208: log likelihood = 18324.722 (backed up)
          flat log likelihood encountered, cannot find uphill direction

          but you are right. the log likelihood looks really flat.

          Yes, I need the conditional variance for each fund, then sort funds into 5 groups based on variance.

          I am trying to use following 2 ways to solve the problem. Hope it will be work.

          first, ever time, when STATA get stuck in converge, the variance still can be predicted for the portfolio of prior id. So, I skip that id portfolio and start again from the next one. Then, I will back to try different GARCH-type model to estimate variance of skipped portfolios. I guess it could be a way, because I tried that use GARCH(1,1) and EGARCH(1,1) for the same id portfolio, there is no big different between conditional variance.

          second, because after daily conditional variances are estimated, I will average the daily variance within each month to get monthly variance. So, I use monthly data to estimate monthly conditional variance. Before, I am worry about the observations. the maximum observation for each fund is 288. The minimum observation for some funds could be 36. I am not sure it works or not, cause I am trying different GARCH family and have not find a suitable one.



          Comment


          • #6
            Originally posted by Yanyu Li View Post
            Dear Weiwen,

            I am really appreciate for your reply.

            actually, I use daily data over 26 years. I know GARCH require a large observations to regress, so I even try to use daily data rather than monthly.
            ...
            but you are right. the log likelihood looks really flat.

            Yes, I need the conditional variance for each fund, then sort funds into 5 groups based on variance.

            I am trying to use following 2 ways to solve the problem. Hope it will be work.

            first, ever time, when STATA get stuck in converge, the variance still can be predicted for the portfolio of prior id. So, I skip that id portfolio and start again from the next one. Then, I will back to try different GARCH-type model to estimate variance of skipped portfolios. I guess it could be a way, because I tried that use GARCH(1,1) and EGARCH(1,1) for the same id portfolio, there is no big different between conditional variance.

            second, because after daily conditional variances are estimated, I will average the daily variance within each month to get monthly variance. So, I use monthly data to estimate monthly conditional variance. Before, I am worry about the observations. the maximum observation for each fund is 288. The minimum observation for some funds could be 36. I am not sure it works or not, cause I am trying different GARCH family and have not find a suitable one.


            My mistake, you did say in your initial post that you were using daily data.

            I went and skimmed the manual for the -arch- command, and it does indicate that this model is very prone to convergence problems. Unfortunately, it also says that there is not always a solution.

            It does look like Stata found a maxima as early as iteration 15, but it could not determine that this was a global maximum, and not a local maxima. You could tell it to run 15 iterations and display the gradient of the log-likelihood. It will display that as a matrix, with one entry for each parameter its estimating. Whichever parameters aren't significantly different from 0 are causing Stata trouble for that fund. If you consistently have some parameters causing trouble, then it may be worth it to check if there are any data entry errors. Or it may even be worth it to drop them, if you can justify that substantively.

            Have other researchers done anything similar to what you are doing? If so, are your methods a lot different from theirs? And is there someone who is more familiar with these methods who you can consult? This definitely seems like a very tricky method, given the convergence problems. Another thing to consider is that when you were using daily data, I know that most mutual funds in the US usually don't see daily changes of more than 0.3%. Is there enough variability to estimate the model at all?

            I don't think I can help further, because while I know what a mutual fund is, I have no experience at all with this statistical model.
            Please use the code delimiters to show code and results - use the # button on the formatting toolbar, between the " (double quote) and <> buttons.

            Please use the command -dataex- to show a representative sample of data; it is installed already if you have Stata 14.2 or 15.1, else you can install it by typing

            Code:
            ssc install dataex

            Comment


            • #7
              Dear weiwen,

              Many thanks for your help. Yes, you are right. The characteristic of fund returns could be a problem. the prior study works on stock market and construct simulated portfolio, not the one actively managed by professional investors.

              I will go back to my original data, trying to find some statistical method to solve this problem. I am really appreciate for everything.

              Comment

              Working...
              X