Dear Statalist,
I am building a panel database (yearly frequency) related to renewable energies production in several countries. This is an example of what it looks like (for the year 2014):
I am performing a Panel VAR analysis, because the objective is to forecast each of the variables in columns for the following 5-10 years for each country. For this, I am using the command 'pvar' (any other suggestions?). One of the results looks like the following:
My questions are the following:
1.) How do I produce forecasts for the following years of the endogenous variables of the model for each country? (I have seen in the command's help file that it can produce Impulse-Response Functions (IRF) and Forecast Error Variance Decomposition (FEVD), but no simple forecasting).
2.) Is there any criteria similar to the R-squared of Ordinary Least Squares (OLS) that can give me and idea of "how good is my model" for prediction? As it can be seen, the output of the command doesn't seem to produce anything similar.
3.) Should I feed the command the endogenous variables in their stationary form, or is it ok if I input the variables in levels? I am given to understand that the command eliminates fixed effects and uses lags of the endogenous variables as instruments, but I am not sure.
Thanks in advance.
Best regards,
Juan Hernandez
I am building a panel database (yearly frequency) related to renewable energies production in several countries. This is an example of what it looks like (for the year 2014):
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input str23 country int year double(conven_GWh hydro_GWh wind_GWh solar_GWh PW_GWh) "Argentina" 2014 106887.6640625 33135.7 619.5 15.9 141560.46511627905 "Australia" 2014 210412.296875 18421 10252 4857.5 247419.28999999998 "Austria" 2014 15165.2041015625 44836.2 3845.8 785.2 65109.104176 "Azerbaijan" 2014 23138.900390625 1299.7 112.7 2.9 24727.7 "Bangladesh" 2014 54634.3984375 993 5.1 212.4 55845 "Belarus" 2014 34482 121 11 2 34737 "Belgium" 2014 60458.3984375 1461.7 4614 2882.9 72672 "Brazil" 2014 158432.21875 373441.9 12210.4 61.3 590542.117779729 "Bulgaria" 2014 40096.30078125 5162.6 1330.6 1252.5 47485 "Canada" 2014 229040.84375 382574.1 22538 2119.8 648631.7508488516 "Chile" 2014 43239.828125 23098.7 1442.9 489.7 73598.129 "China" 2014 4374041 1060100 158576.5 25514.1 5649583.000000001 "Czech Republic" 2014 76831.796875 2960.7 476.5 2122.9 86003.4 "Denmark" 2014 14200.9306640625 15.1 13076.5 595.5 32181.730555555554 "Ecuador" 2014 12353.61328125 11457.9 79.7 16.5 24307.213342674 "Egypt" 2014 155474 13352 1332 42 170200 "Finland" 2014 41813.69921875 13397 1107.2 7.8 68084 "France" 2014 469878.1875 68627 17249 5913.3 561685.4816458916 "Germany" 2014 464484.59375 25443.9 57357.1 36056.5 626650.1017700001 "Greece" 2014 38297 4607 3689 3792 50474 "Hungary" 2014 26251.599609375 301.5 656.5 56 29392 "India" 2014 1066293.125 134193.1 33454.9 5076 1252028.109529148 "Indonesia" 2014 196404.296875 15148 1.2 10.8 228489 "Iran" 2014 260319.796875 13865.8 358 18.9 274599.99999999994 "Ireland" 2014 19927.904296875 987.7 5140.1 1.1 26318.704845515826 "Israel" 2014 60385.6015625 23.1 8.9 774.3 61295 "Italy" 2014 159137.09375 60256.3 15178.3 22318.8 279828.5 "Japan" 2014 928397.125 87564.1 5190.5 26534 1062728.819620551 "Kazakhstan" 2014 86288.1015625 8262.8 13.3 1.3 94567.4 "Lithuania" 2014 2886.89990234375 1087 639 73 4396.9 "Mexico" 2014 250422.828125 38892.8 6426.2 220.7 303315.8327426527 "Netherlands" 2014 91656.6015625 112.2 5797.3 784.8 103365 "New Zealand" 2014 9142.091796875 24316.5 2214.1 17 43598.6916006103 "Norway" 2014 4040.199951171875 136183 2216 10.7 141967 "Pakistan" 2014 71890.2890625 33201 459.3 241.1 107159.1866828133 "Peru" 2014 22656.19921875 22201.3 257.5 199.3 45549.8 "Philippines" 2014 57451.49609375 9811.7 152.1 16.5 77260.997 "Poland" 2014 139216.203125 2733.8 7675.6 6.9 159058 "Portugal" 2014 21234.828125 16411.9 12111 627.3 52802.12783900001 "Romania" 2014 36158.8984375 19279.1 6200.9 1616 63284.7 "Russian Federation" 2014 884480.1875 175595 5 7.1 1058700.9000000001 "Slovakia" 2014 21025 4462 6 597 27254 "South_Africa" 2014 251134 3957.9 1056.9 1189 254663 "South_Korea" 2014 524609.8125 14464 1169.5 2556.3 540378.799 "Spain" 2014 168480.03125 42971 52013 13673 278750.1245269591 "Sweden" 2014 67921.8984375 63871.4 11234 47 153662 "Switzerland" 2014 34204.1953125 39701 101 842 74874.19354838714 "Taiwan - Chinese Taipei" 2014 251652.46875 7439.1 1500.5 551.7 259975.06199999998 "Thailand" 2014 151335.28125 6017.8 305.1 1564.3 173764.180391 "Turkey" 2014 199725.6875 40253 8520 20.1 251962.78 "Ukraine" 2014 172535.703125 9291 1171.5 482.5 182815 "United_Kingdom" 2014 274303.3125 8775.9 31965.9 4039.8 338174.60774709244 "United_States" 2014 3803829.75 281527.2 183891.8 24603.1 4363326.4072524905 "Venezuela" 2014 23995.400390625 86322 88.3 5.3 110411 end
Code:
pvar PW_GWh_2 conven_GWh_2 hydro_GWh_2 wind_GWh_2 solar_GWh_2 Panel vector autoregresssion GMM Estimation Final GMM Criterion Q(b) = 1.04e-27 Initial weight matrix: Identity GMM weight matrix: Robust No. of obs = 542 No. of panels = 54 Ave. no. of T = 10.037 ------------------------------------------------------------------------------ | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- PW_GWh_2 | PW_GWh_2 | L1. | 3.06515 4.463189 0.69 0.492 -5.682539 11.81284 | conven_GWh_2 | L1. | -1.765701 4.411304 -0.40 0.689 -10.4117 6.880296 | hydro_GWh_2 | L1. | -3.286836 4.943308 -0.66 0.506 -12.97554 6.40187 | wind_GWh_2 | L1. | -2.813435 5.188359 -0.54 0.588 -12.98243 7.355562 | solar_GWh_2 | L1. | -.9791303 6.151945 -0.16 0.874 -13.03672 11.07846 -------------+---------------------------------------------------------------- conven_GWh_2 | PW_GWh_2 | L1. | .3722461 4.69683 0.08 0.937 -8.833371 9.577863 | conven_GWh_2 | L1. | .8286457 4.565836 0.18 0.856 -8.120228 9.77752 | hydro_GWh_2 | L1. | -1.302162 5.524593 -0.24 0.814 -12.13017 9.525841 | wind_GWh_2 | L1. | -1.033817 5.518745 -0.19 0.851 -11.85036 9.782725 | solar_GWh_2 | L1. | 1.667345 6.033538 0.28 0.782 -10.15817 13.49286 -------------+---------------------------------------------------------------- hydro_GWh_2 | PW_GWh_2 | L1. | -.3210172 1.969008 -0.16 0.870 -4.180202 3.538167 | conven_GWh_2 | L1. | .3673109 1.949041 0.19 0.851 -3.452739 4.187361 | hydro_GWh_2 | L1. | 1.231796 2.177828 0.57 0.572 -3.036668 5.500261 | wind_GWh_2 | L1. | .3494463 2.218587 0.16 0.875 -3.998904 4.697796 | solar_GWh_2 | L1. | -.0423122 2.786329 -0.02 0.988 -5.503416 5.418792 -------------+---------------------------------------------------------------- wind_GWh_2 | PW_GWh_2 | L1. | 2.451475 1.32013 1.86 0.063 -.1359313 5.038881 | conven_GWh_2 | L1. | -2.394254 1.309703 -1.83 0.068 -4.961224 .1727166 | hydro_GWh_2 | L1. | -2.689287 1.432047 -1.88 0.060 -5.496047 .117473 | wind_GWh_2 | L1. | -1.72751 1.494309 -1.16 0.248 -4.656301 1.201281 | solar_GWh_2 | L1. | -2.940695 1.842666 -1.60 0.111 -6.552255 .6708651 -------------+---------------------------------------------------------------- solar_GWh_2 | PW_GWh_2 | L1. | -.437096 .254252 -1.72 0.086 -.9354207 .0612287 | conven_GWh_2 | L1. | .4334507 .2500042 1.73 0.083 -.0565486 .9234499 | hydro_GWh_2 | L1. | .4638522 .2837395 1.63 0.102 -.0922669 1.019971 | wind_GWh_2 | L1. | .5768785 .2873042 2.01 0.045 .0137726 1.139984 | solar_GWh_2 | L1. | 1.484157 .337831 4.39 0.000 .8220208 2.146294 ------------------------------------------------------------------------------ Instruments : l(1/1).(PW_GWh_2 conven_GWh_2 hydro_GWh_2 wind_GWh_2 solar_GWh_2)
1.) How do I produce forecasts for the following years of the endogenous variables of the model for each country? (I have seen in the command's help file that it can produce Impulse-Response Functions (IRF) and Forecast Error Variance Decomposition (FEVD), but no simple forecasting).
2.) Is there any criteria similar to the R-squared of Ordinary Least Squares (OLS) that can give me and idea of "how good is my model" for prediction? As it can be seen, the output of the command doesn't seem to produce anything similar.
3.) Should I feed the command the endogenous variables in their stationary form, or is it ok if I input the variables in levels? I am given to understand that the command eliminates fixed effects and uses lags of the endogenous variables as instruments, but I am not sure.
Thanks in advance.
Best regards,
Juan Hernandez
Comment