Hello,
I have fishery data by vessels (i.e. firms) and trips ( i.e. trips made to the sea to catch fishes). In this sample, there are 8 vessels, with a total of 1261 trips. These trips took place between 2001 and 2019 for a total of 19 years. In a given year a vessel (i.e., a firm) can make several trips.
Each vessel represents a firm, and the vessel variable is named "Firm" in my dataset. The dataset is identified by two key variables: "Firm" and "Trip."
In this dataset, I have information about the prices of three different fish species caught by these vessels written as price_1, price_2, and price_3, capital used for each trip registered as capital and Year variable, and the total revenue gained from selling these three fish species for each vessel and trip.
All the variables vary by vessel ( firm) and Trip.
I have used the "dataex" command to display the first 30 observations of this sample dataset. Below you can see the variables of interest for the first 30 observations of my dataset.
dataex Firm Trip year Capital Price_1 Price_2 Price_3 TotalRevenue Totalrevenue_div_price_1 ln_Totalrevenue_div_price_1 in 1/30
----------------------- copy starting from the next line -----------------------
[CODE]
* Example generated by -dataex-. For more info, type help dataex
clear
input long Firm float(Trip year) int Capital double Price_1 float Price_2 double(Price_3 TotalRevenue) float(Totalrevenue_div_price_1 ln_Totalrevenue_div_price_1)
Now, my question is regarding a problem I have encountered while running a specific model. The issue is about missing p-values, confidence intervals, and t-values for some parameters of the regression results.
The regression I am running is a translog revenue function, which assesses output-oriented technical efficiency by incorporating firm fixed effect parameters using firm dummies. Additionally, it includes two price distortion parameters to measure allocation biases. This modelling approach aligns is in line with Kumbhakar and Lovell (2003) and Asche and Roll (2018). I used the STATA nl command for non-linear models.
To run the model, have to use initial values for the two price distortion parameters. I am sharing two of the results from several initial value trials, chosen randomly.
The challenge I am facing is that in all my results, I observe missing p-values, confidence intervals, and t-values for two or three parameters. Furthermore, the parameter experiencing this issue changes when I alter the initial values.
I am very much interested in your insights on how to address this problem of the absence of p-values, confidence intervals, and t-values in the parameters of the estimated results. Do you have any tips on why this might be happening? I have already checked the correlation between variables used in my regression model. None of them show a high correlation. The highest correlation I found is 0.6750, which is between capital and total revenue, representing the dependent variable and one of the independent variables. The lowest correlation is 0.0751, which exists between price 1 and price 3. Hence, I become less concerned that correlation is causing this.
I presented to you below the screenshot of two of my results
e.g 1) When I use initial values of (θBA 0.01 θCA 0.01) for both of the price bias parameters, the coefficient of firm 4, the coefficient of price_3 and the coefficient of capital get missing p-values, confidence intervals, and t-values.
nl (ln_Totalrevenue_div_price_1 = {b1}*Firm1 + {b2}*Firm2 + {b3}*Firm3 + {b4}*Firm4 + {b5}*Firm5 + {b6}*Firm6 + {b7}*Firm7 + {b8}*Firm8 + {dB}*ln(Price_2/Price_1) + {dC}*ln(Price_3/Price_1) + 1/2*{dBB}*ln(Price_2/Price_1)*ln(Price_2/Price_1) + 1/2*{dCC}*ln(Price_3/Price_1) * ln(Price_3/Price_1) + {dBC}*ln(Price_2/Price_1)* ln(Price_3/Price_1) + {dK}*ln(Capital) + 1/2*{dKK}*ln(Capital)*ln(Capital) + {dBK}*ln(Price_2/Price_1)*ln(Capital) + {dCK}*ln(Price_3/Price_1)*ln(Capital) + {dY}*year + 1/2*{dYY}*year*year + ( {dB}*ln({θBA}) + {dBC}*ln({θCA})*ln(Price_2/Price_1) + {dBK}*ln(Capital)*ln({θBA}) + {dC}*ln({θCA}) + {dBC}*ln({θBA})*ln(Price_3/Price_1) + {dCK}*ln(Capital)*ln({θCA}) + (1/2)*(({dBC}*ln({θCA})*ln({θBA})) + ({dBC}*ln({θCA})*ln({θBA}))))) , noconstant initial (θBA 0.01 θCA 0.01)
e.g. 2) When I use initial values (θBA 0.01 θCA 1.01 ) for both of the price bias parameters, the price distortion parameters of the price of product 3 relative to the price of product 1 ( that is θCA ) gets missing p-values, confidence intervals, and t-values.
nl (ln_Totalrevenue_div_price_1 = {b1}*Firm1 + {b2}*Firm2 + {b3}*Firm3 + {b4}*Firm4 + {b5}*Firm5 + {b6}*Firm6 + {b7}*Firm7 + {b8}*Firm8 + {dB}*ln(Price_2/Price_1) + {dC}*ln(Price_3/Price_1) + 1/2*{dBB}*ln(Price_2/Price_1)*ln(Price_2/Price_1) + 1/2*{dCC}*ln(Price_3/Price_1) * ln(Price_3/Price_1) + {dBC}*ln(Price_2/Price_1)* ln(Price_3/Price_1) + {dK}*ln(Capital) + 1/2*{dKK}*ln(Capital)*ln(Capital) + {dBK}*ln(Price_2/Price_1)*ln(Capital) + {dCK}*ln(Price_3/Price_1)*ln(Capital) + {dY}*year + 1/2*{dYY}*year*year + ( {dB}*ln({θBA}) + {dBC}*ln({θCA})*ln(Price_2/Price_1) + {dBK}*ln(Capital)*ln({θBA}) + {dC}*ln({θCA}) + {dBC}*ln({θBA})*ln(Price_3/Price_1) + {dCK}*ln(Capital)*ln({θCA}) + (1/2)*(({dBC}*ln({θCA})*ln({θBA})) + ({dBC}*ln({θCA})*ln({θBA}))))) , noconstant initial (θBA 0.01 θCA 1.01)
I have fishery data by vessels (i.e. firms) and trips ( i.e. trips made to the sea to catch fishes). In this sample, there are 8 vessels, with a total of 1261 trips. These trips took place between 2001 and 2019 for a total of 19 years. In a given year a vessel (i.e., a firm) can make several trips.
Each vessel represents a firm, and the vessel variable is named "Firm" in my dataset. The dataset is identified by two key variables: "Firm" and "Trip."
In this dataset, I have information about the prices of three different fish species caught by these vessels written as price_1, price_2, and price_3, capital used for each trip registered as capital and Year variable, and the total revenue gained from selling these three fish species for each vessel and trip.
All the variables vary by vessel ( firm) and Trip.
I have used the "dataex" command to display the first 30 observations of this sample dataset. Below you can see the variables of interest for the first 30 observations of my dataset.
dataex Firm Trip year Capital Price_1 Price_2 Price_3 TotalRevenue Totalrevenue_div_price_1 ln_Totalrevenue_div_price_1 in 1/30
----------------------- copy starting from the next line -----------------------
[CODE]
* Example generated by -dataex-. For more info, type help dataex
clear
input long Firm float(Trip year) int Capital double Price_1 float Price_2 double(Price_3 TotalRevenue) float(Totalrevenue_div_price_1 ln_Totalrevenue_div_price_1)
| 1990004501 | 257 | 2001 | 1559 | 5.469641916437395 | 10.923235 | 1.3090858459472656 | 7877810.152981914 | 1440279 | 14.180347 |
| 1990004501 | 265 | 2001 | 1559 | 5.469641916437395 | 10.311555 | 1.3090858459472656 | 8087359.777981914 | 1478590.4 | 14.2066 |
| 1990004501 | 4 | 2001 | 1559 | 3.8147547341003065 | 10.377046 | 1.3090858459472656 | 8223249.563577672 | 2155643 | 14.5836 |
| 1990004501 | 273 | 2001 | 1559 | 5.469641916437395 | 10.586683 | 1.3090858459472656 | 7784215.527981914 | 1423167.3 | 14.168395 |
| 1990004501 | 80 | 2001 | 1559 | 5.469641916437395 | 10.377046 | 1.2513922717778807 | 7513275.849919865 | 1373632 | 14.13297 |
| 1990004501 | 160 | 2001 | 1559 | 5.469641916437395 | 10.377046 | 1.3090857881949458 | 7087651.027981914 | 1295816.3 | 14.074652 |
| 1987003408 | 277 | 2001 | 2181 | 6.145504215642849 | 10.513448 | 1.2102010250091553 | 8831620.131975446 | 1437086.4 | 14.178128 |
| 1987003408 | 58 | 2001 | 2181 | 6.145504215642849 | 10.3646 | 1.2114313752096442 | 8927751.182878688 | 1452729 | 14.188954 |
| 1987003408 | 140 | 2001 | 2181 | 6.145504215642849 | 10.3646 | 1.2102010686355984 | 8319637.381975447 | 1353776.3 | 14.118408 |
| 1990004501 | 40 | 2001 | 1559 | 5.469641916437395 | 10.377046 | 1.310155227830855 | 7281510.755808509 | 1331259 | 14.101636 |
| 1987003408 | 102 | 2001 | 2181 | 6.145504215642849 | 10.3646 | 1.144262752428504 | 8711770.586418048 | 1417584.4 | 14.164465 |
| 1987003408 | 30 | 2001 | 2181 | 6.145504215642849 | 10.3646 | 1.1935991754736714 | 8241539.064098619 | 1341068 | 14.108977 |
| 1987003408 | 82 | 2001 | 2181 | 6.145504215642849 | 10.3646 | 1.1601555762542333 | 8718536.306050256 | 1418685.3 | 14.16524 |
| 1979006694 | 84 | 2001 | 458 | 5.221693527642477 | 9.23322 | 3.3233383622128017 | 3836800.407281849 | 734780.9 | 13.507328 |
| 1990004501 | 60 | 2001 | 1559 | 5.469641916437395 | 10.377046 | 1.2812073987762707 | 7687143.22158874 | 1405420 | 14.155847 |
| 1979006694 | 66 | 2001 | 458 | 5.221693527642477 | 9.23322 | 4.603591692273926 | 3996239.2853910658 | 765314.8 | 13.548042 |
| 1987003408 | 146 | 2001 | 2181 | 6.011194541023049 | 10.3646 | 1.2102010250091553 | 8950989.906164631 | 1489053.5 | 14.213652 |
| 1987003408 | 41 | 2001 | 2181 | 6.145504215642849 | 10.3646 | 1.2907081870606796 | 8500836.39537295 | 1383261 | 14.139955 |
| 1990004501 | 72 | 2001 | 1559 | 5.469641916437395 | 10.377046 | 1.2605326621060873 | 7640762.367082128 | 1396940 | 14.149795 |
| 1990004501 | 263 | 2001 | 1559 | 5.469641916437395 | 10.449423 | 1.3090858459472656 | 7811639.027981914 | 1428181 | 14.171912 |
| 1990004501 | 171 | 2001 | 1559 | 5.535629421373175 | 10.377046 | 1.3090858459472656 | 7922497.462540114 | 1431182.8 | 14.174012 |
| 1987003408 | 305 | 2001 | 2181 | 6.145504215642849 | 10.3646 | 1.2102010686355984 | 8319637.381975447 | 1353776.3 | 14.118408 |
| 1990004501 | 313 | 2001 | 1559 | 5.469641916437395 | 10.377046 | 1.3090857881949458 | 7087651.027981914 | 1295816.3 | 14.074652 |
| 1990004501 | 276 | 2001 | 1559 | 5.7527606403274305 | 10.377046 | 1.3090858459472656 | 7309688.284373303 | 1270640 | 14.055032 |
| 1990004501 | 302 | 2001 | 1559 | 5.469641916437395 | 10.377046 | 1.3090857881949458 | 7087651.027981914 | 1295816.3 | 14.074652 |
| 1990004501 | 329 | 2001 | 1559 | 5.860018978756407 | 10.377046 | 1.3090858459472656 | 8113189.134433766 | 1384498.8 | 14.14085 |
| 1990004501 | 305 | 2001 | 1559 | 5.469641916437395 | 10.377046 | 1.3090857881949458 | 7087651.027981914 | 1295816.3 | 14.074652 |
| 1987003408 | 302 | 2001 | 2181 | 6.145504215642849 | 10.3646 | 1.2102010686355984 | 8319637.381975447 | 1353776.3 | 14.118408 |
| 1990004501 | 166 | 2001 | 1559 | 5.388732750914947 | 10.377046 | 1.3090858459472656 | 7463122.963566745 | 1384949.5 | 14.141174 |
| 1987003408 | 170 | 2001 | 2181 | 6.145504215642849 | 10.3646 | 1.0904509792301724 | 7601022.096657071 | 1236842.8 | 14.028072 |
The regression I am running is a translog revenue function, which assesses output-oriented technical efficiency by incorporating firm fixed effect parameters using firm dummies. Additionally, it includes two price distortion parameters to measure allocation biases. This modelling approach aligns is in line with Kumbhakar and Lovell (2003) and Asche and Roll (2018). I used the STATA nl command for non-linear models.
To run the model, have to use initial values for the two price distortion parameters. I am sharing two of the results from several initial value trials, chosen randomly.
The challenge I am facing is that in all my results, I observe missing p-values, confidence intervals, and t-values for two or three parameters. Furthermore, the parameter experiencing this issue changes when I alter the initial values.
I am very much interested in your insights on how to address this problem of the absence of p-values, confidence intervals, and t-values in the parameters of the estimated results. Do you have any tips on why this might be happening? I have already checked the correlation between variables used in my regression model. None of them show a high correlation. The highest correlation I found is 0.6750, which is between capital and total revenue, representing the dependent variable and one of the independent variables. The lowest correlation is 0.0751, which exists between price 1 and price 3. Hence, I become less concerned that correlation is causing this.
I presented to you below the screenshot of two of my results
e.g 1) When I use initial values of (θBA 0.01 θCA 0.01) for both of the price bias parameters, the coefficient of firm 4, the coefficient of price_3 and the coefficient of capital get missing p-values, confidence intervals, and t-values.
nl (ln_Totalrevenue_div_price_1 = {b1}*Firm1 + {b2}*Firm2 + {b3}*Firm3 + {b4}*Firm4 + {b5}*Firm5 + {b6}*Firm6 + {b7}*Firm7 + {b8}*Firm8 + {dB}*ln(Price_2/Price_1) + {dC}*ln(Price_3/Price_1) + 1/2*{dBB}*ln(Price_2/Price_1)*ln(Price_2/Price_1) + 1/2*{dCC}*ln(Price_3/Price_1) * ln(Price_3/Price_1) + {dBC}*ln(Price_2/Price_1)* ln(Price_3/Price_1) + {dK}*ln(Capital) + 1/2*{dKK}*ln(Capital)*ln(Capital) + {dBK}*ln(Price_2/Price_1)*ln(Capital) + {dCK}*ln(Price_3/Price_1)*ln(Capital) + {dY}*year + 1/2*{dYY}*year*year + ( {dB}*ln({θBA}) + {dBC}*ln({θCA})*ln(Price_2/Price_1) + {dBK}*ln(Capital)*ln({θBA}) + {dC}*ln({θCA}) + {dBC}*ln({θBA})*ln(Price_3/Price_1) + {dCK}*ln(Capital)*ln({θCA}) + (1/2)*(({dBC}*ln({θCA})*ln({θBA})) + ({dBC}*ln({θCA})*ln({θBA}))))) , noconstant initial (θBA 0.01 θCA 0.01)
e.g. 2) When I use initial values (θBA 0.01 θCA 1.01 ) for both of the price bias parameters, the price distortion parameters of the price of product 3 relative to the price of product 1 ( that is θCA ) gets missing p-values, confidence intervals, and t-values.
nl (ln_Totalrevenue_div_price_1 = {b1}*Firm1 + {b2}*Firm2 + {b3}*Firm3 + {b4}*Firm4 + {b5}*Firm5 + {b6}*Firm6 + {b7}*Firm7 + {b8}*Firm8 + {dB}*ln(Price_2/Price_1) + {dC}*ln(Price_3/Price_1) + 1/2*{dBB}*ln(Price_2/Price_1)*ln(Price_2/Price_1) + 1/2*{dCC}*ln(Price_3/Price_1) * ln(Price_3/Price_1) + {dBC}*ln(Price_2/Price_1)* ln(Price_3/Price_1) + {dK}*ln(Capital) + 1/2*{dKK}*ln(Capital)*ln(Capital) + {dBK}*ln(Price_2/Price_1)*ln(Capital) + {dCK}*ln(Price_3/Price_1)*ln(Capital) + {dY}*year + 1/2*{dYY}*year*year + ( {dB}*ln({θBA}) + {dBC}*ln({θCA})*ln(Price_2/Price_1) + {dBK}*ln(Capital)*ln({θBA}) + {dC}*ln({θCA}) + {dBC}*ln({θBA})*ln(Price_3/Price_1) + {dCK}*ln(Capital)*ln({θCA}) + (1/2)*(({dBC}*ln({θCA})*ln({θBA})) + ({dBC}*ln({θCA})*ln({θBA}))))) , noconstant initial (θBA 0.01 θCA 1.01)
