------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- name: log: F:\vhm812-data\l3a_model_build_II.log log type: text opened on: 19 Jan 2015, 13:38:05 . set more off . . * open the DAISY Red dataset . use daisy2red.dta, clear . gen month=month(calv_dt) . gen aut_calv=(month>=2 & month<=7) . gen hs_ct=herd_size-251 . gen hs_sq=herd_size^2 . gen parity1=parity-1 . gen milk120k=milk120/1000 (38 missing values generated) . gen wpc_sqrt=sqrt(wpc) . . * specifying maximum model . * no analyses . . * causal model . * no analyses . . * Reducing the Number of Predictors . * descriptive statistics . codebook cf vag_disch ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- cf Calving to first service interval ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- type: numeric (float) range: [21,240] units: 1 unique values: 129 missing .: 14/1574 mean: 71.3115 std. dev: 21.9039 percentiles: 10% 25% 50% 75% 90% 51 59 67 78 96 ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- vag_disch Vaginal discharge observed ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- type: numeric (float) label: noyes range: [0,1] units: 1 unique values: 2 missing .: 0/1574 tabulation: Freq. Numeric Label 1492 0 no 82 1 yes . sum milk120 parity herd_size dyst rp vag_disch, detail Milk volume (l) in first 120 days of lactation ------------------------------------------------------------- Percentiles Smallest 1% 1698.5 1110.2 5% 2077.2 1397.7 10% 2298.2 1461.5 Obs 1536 25% 2731.95 1467 Sum of Wgt. 1536 50% 3215.25 Mean 3215.096 Largest Std. Dev. 698.1316 75% 3682.1 5181.8 90% 4080.7 5278 Variance 487387.7 95% 4403.3 5399.9 Skewness .1101838 99% 4904.4 5630.3 Kurtosis 2.845637 Lactation number ------------------------------------------------------------- Percentiles Smallest 1% 1 1 5% 1 1 10% 1 1 Obs 1574 25% 1 1 Sum of Wgt. 1574 50% 2 Mean 2.729987 Largest Std. Dev. 1.493841 75% 4 7 90% 5 7 Variance 2.23156 95% 5 7 Skewness .5450922 99% 6 7 Kurtosis 2.315593 Herd size ------------------------------------------------------------- Percentiles Smallest 1% 125 125 5% 125 125 10% 185 125 Obs 1574 25% 201 125 Sum of Wgt. 1574 50% 263 Mean 251.0076 Largest Std. Dev. 62.01692 75% 294 333 90% 333 333 Variance 3846.098 95% 333 333 Skewness -.3550929 99% 333 333 Kurtosis 2.256969 Dystocia at calving ------------------------------------------------------------- Percentiles Smallest 1% 0 0 5% 0 0 10% 0 0 Obs 1574 25% 0 0 Sum of Wgt. 1574 50% 0 Mean .0597205 Largest Std. Dev. .2370435 75% 0 1 90% 0 1 Variance .0561896 95% 1 1 Skewness 3.715938 99% 1 1 Kurtosis 14.80819 Retained placenta at calving ------------------------------------------------------------- Percentiles Smallest 1% 0 0 5% 0 0 10% 0 0 Obs 1574 25% 0 0 Sum of Wgt. 1574 50% 0 Mean .0946633 Largest Std. Dev. .2928423 75% 0 1 90% 0 1 Variance .0857566 95% 1 1 Skewness 2.769173 99% 1 1 Kurtosis 8.66832 Vaginal discharge observed ------------------------------------------------------------- Percentiles Smallest 1% 0 0 5% 0 0 10% 0 0 Obs 1574 25% 0 0 Sum of Wgt. 1574 50% 0 Mean .0520966 Largest Std. Dev. .2222924 75% 0 1 90% 0 1 Variance .0494139 95% 1 1 Skewness 4.031139 99% 1 1 Kurtosis 17.25008 . . * correlation . corr milk120 parity herd_size (obs=1536) | milk120 parity herd_s~e -------------+--------------------------- milk120 | 1.0000 parity | 0.3821 1.0000 herd_size | -0.0433 0.0356 1.0000 . pwcorr milk120 parity herd_size, obs star(0.05) | milk120 parity herd_s~e -------------+--------------------------- milk120 | 1.0000 | 1536 | parity | 0.3821* 1.0000 | 1536 1574 | herd_size | -0.0433 0.0386 1.0000 | 1536 1574 1574 | . . * indices . * no analyses . . * unconditional associations . reg cf parity Source | SS df MS Number of obs = 1560 -------------+------------------------------ F( 1, 1558) = 1.46 Model | 699.698494 1 699.698494 Prob > F = 0.2273 Residual | 747280.894 1558 479.641139 R-squared = 0.0009 -------------+------------------------------ Adj R-squared = 0.0003 Total | 747980.592 1559 479.782291 Root MSE = 21.901 ------------------------------------------------------------------------------ cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- parity | -.4480144 .3709323 -1.21 0.227 -1.175594 .2795649 _cons | 72.53554 1.155186 62.79 0.000 70.26965 74.80142 ------------------------------------------------------------------------------ . reg cf vag_disch Source | SS df MS Number of obs = 1560 -------------+------------------------------ F( 1, 1558) = 0.49 Model | 236.166881 1 236.166881 Prob > F = 0.4831 Residual | 747744.425 1558 479.938656 R-squared = 0.0003 -------------+------------------------------ Adj R-squared = -0.0003 Total | 747980.592 1559 479.782291 Root MSE = 21.908 ------------------------------------------------------------------------------ cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- vag_disch | 1.743523 2.485484 0.70 0.483 -3.131724 6.61877 _cons | 71.21989 .5698436 124.98 0.000 70.10215 72.33763 ------------------------------------------------------------------------------ . . * principle components / factor analysis / correspondence analysis . * not covered . . * Functional Form of predictors . * residual plots . reg cf milk120 Source | SS df MS Number of obs = 1525 -------------+------------------------------ F( 1, 1523) = 0.72 Model | 329.282002 1 329.282002 Prob > F = 0.3977 Residual | 700869.964 1523 460.19039 R-squared = 0.0005 -------------+------------------------------ Adj R-squared = -0.0002 Total | 701199.246 1524 460.104492 Root MSE = 21.452 ------------------------------------------------------------------------------ cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- milk120 | .0006667 .0007881 0.85 0.398 -.0008793 .0022126 _cons | 68.95612 2.595207 26.57 0.000 63.86556 74.04667 ------------------------------------------------------------------------------ . predict stdres, rstandar (49 missing values generated) . . * lowess smoother . twoway (scatter stdres milk120) (lowess stdres milk120) . . * Detecting and Correcting for non-linearity (transformation of X) . * categorization of predictor . reg cf parity Source | SS df MS Number of obs = 1560 -------------+------------------------------ F( 1, 1558) = 1.46 Model | 699.698494 1 699.698494 Prob > F = 0.2273 Residual | 747280.894 1558 479.641139 R-squared = 0.0009 -------------+------------------------------ Adj R-squared = 0.0003 Total | 747980.592 1559 479.782291 Root MSE = 21.901 ------------------------------------------------------------------------------ cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- parity | -.4480144 .3709323 -1.21 0.227 -1.175594 .2795649 _cons | 72.53554 1.155186 62.79 0.000 70.26965 74.80142 ------------------------------------------------------------------------------ . egen parity_c6=cut(parity), at(0 1 2 3 4 5 6 15) icodes . tab parity parity_c6 Lactation | parity_c6 number | 1 2 3 4 5 6 | Total -----------+------------------------------------------------------------------+---------- 1 | 417 0 0 0 0 0 | 417 2 | 0 374 0 0 0 0 | 374 3 | 0 0 319 0 0 0 | 319 4 | 0 0 0 222 0 0 | 222 5 | 0 0 0 0 169 0 | 169 6 | 0 0 0 0 0 69 | 69 7 | 0 0 0 0 0 4 | 4 -----------+------------------------------------------------------------------+---------- Total | 417 374 319 222 169 73 | 1,574 . reg cf i.parity_c6 Source | SS df MS Number of obs = 1560 -------------+------------------------------ F( 5, 1554) = 0.91 Model | 2174.44893 5 434.889785 Prob > F = 0.4761 Residual | 745806.143 1554 479.926733 R-squared = 0.0029 -------------+------------------------------ Adj R-squared = -0.0003 Total | 747980.592 1559 479.782291 Root MSE = 21.907 ------------------------------------------------------------------------------ cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- parity_c6 | 2 | -.6468817 1.568168 -0.41 0.680 -3.722829 2.429066 3 | -3.181369 1.635853 -1.94 0.052 -6.390081 .027343 4 | -1.528531 1.831255 -0.83 0.404 -5.120523 2.063462 5 | -2.322236 2.004684 -1.16 0.247 -6.254406 1.609935 6 | -.9349232 2.781437 -0.34 0.737 -6.390688 4.520841 | _cons | 72.61985 1.077984 67.37 0.000 70.5054 74.73431 ------------------------------------------------------------------------------ . . * quadratic function of X . gen ln_cf=ln(cf) (14 missing values generated) . reg ln_cf c.milk120k##c.milk120k Source | SS df MS Number of obs = 1525 -------------+------------------------------ F( 2, 1522) = 4.78 Model | .709133443 2 .354566722 Prob > F = 0.0085 Residual | 112.944101 1522 .074207688 R-squared = 0.0062 -------------+------------------------------ Adj R-squared = 0.0049 Total | 113.653234 1524 .074575613 Root MSE = .27241 --------------------------------------------------------------------------------------- ln_cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] ----------------------+---------------------------------------------------------------- milk120k | .2056932 .0697547 2.95 0.003 .0688677 .3425186 | c.milk120k#c.milk120k | -.0295122 .0105961 -2.79 0.005 -.0502966 -.0087277 | _cons | 3.883482 .1122207 34.61 0.000 3.663358 4.103605 --------------------------------------------------------------------------------------- . estat vif Variable | VIF 1/VIF -------------+---------------------- milk120k | 48.58 0.020585 c.milk120k#| c.milk120k | 48.58 0.020585 -------------+---------------------- Mean VIF | 48.58 . vce, corr Correlation matrix of coefficients of regress model | c.m~120k# e(V) | milk120k c.m~120k _cons -------------+------------------------------ milk120k | 1.0000 c.milk120k#| c.milk120k | -0.9897 1.0000 _cons | -0.9872 0.9559 1.0000 . . * redoing the analysis with milk120 centred . summ milk120k Variable | Obs Mean Std. Dev. Min Max -------------+-------------------------------------------------------- milk120k | 1536 3.215096 .6981316 1.1102 5.6303 . gen m120k_ct=(milk120k-r(mean)) /*r(mean) - memory variable created by summ command that store the mean of the variable*/ (38 missing values generated) . reg ln_cf c.m120k_ct##c.m120k_ct Source | SS df MS Number of obs = 1525 -------------+------------------------------ F( 2, 1522) = 4.78 Model | .709133424 2 .354566712 Prob > F = 0.0085 Residual | 112.944101 1522 .074207688 R-squared = 0.0062 -------------+------------------------------ Adj R-squared = 0.0049 Total | 113.653234 1524 .074575613 Root MSE = .27241 --------------------------------------------------------------------------------------- ln_cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] ----------------------+---------------------------------------------------------------- m120k_ct | .0159242 .0100484 1.58 0.113 -.0037859 .0356344 | c.m120k_ct#c.m120k_ct | -.0295122 .0105961 -2.79 0.005 -.0502966 -.0087277 | _cons | 4.239742 .0086679 489.13 0.000 4.22274 4.256745 --------------------------------------------------------------------------------------- . estat vif Variable | VIF 1/VIF -------------+---------------------- m120k_ct | 1.01 0.992010 c.m120k_ct#| c.m120k_ct | 1.01 0.992010 -------------+---------------------- Mean VIF | 1.01 . vce, corr Correlation matrix of coefficients of regress model | c.m120~t# e(V) | m120k_ct c.m120~t _cons -------------+------------------------------ m120k_ct | 1.0000 c.m120k_ct#| c.m120k_ct | -0.0894 1.0000 _cons | 0.0494 -0.5936 1.0000 . . capture drop stdres . predict stdres, rstandar (49 missing values generated) . twoway (scatter stdres m120k_ct) (lowess stdres m120k_ct) . . *box cox . boxcox ln_cf milk120k , model(rhs) Fitting full model Iteration 0: log likelihood = -183.07949 (not concave) Iteration 1: log likelihood = -180.48708 Iteration 2: log likelihood = -180.32784 Iteration 3: log likelihood = -180.1167 Iteration 4: log likelihood = -179.88238 Iteration 5: log likelihood = -179.86958 Iteration 6: log likelihood = -179.86958 (38 missing values generated) (38 missing values generated) (38 missing values generated) Number of obs = 1525 LR chi2(2) = 8.21 Log likelihood = -179.86958 Prob > chi2 = 0.016 ------------------------------------------------------------------------------ ln_cf | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- /lambda | -3.647194 1.174882 -3.10 0.002 -5.94992 -1.344468 ------------------------------------------------------------------------------ Estimates of scale-variant parameters ---------------------------- | Coef. -------------+-------------- Notrans | _cons | 3.60132 -------------+-------------- Trans | milk120k | 2.329791 -------------+-------------- /sigma | .2722618 ---------------------------- --------------------------------------------------------- Test Restricted LR statistic P-value H0: log likelihood chi2 Prob > chi2 --------------------------------------------------------- lambda = -1 -181.5923 3.45 0.063 lambda = 0 -182.42186 5.10 0.024 lambda = 1 -183.07949 6.42 0.011 --------------------------------------------------------- . end of do-file . do "C:\Users\javier\AppData\Local\Temp\STD01000000.tmp" . gen inv_m120k=1/milk120k (38 missing values generated) . . reg ln_cf inv_m120k Source | SS df MS Number of obs = 1525 -------------+------------------------------ F( 1, 1523) = 4.77 Model | .354674587 1 .354674587 Prob > F = 0.0292 Residual | 113.29856 1523 .0743917 R-squared = 0.0031 -------------+------------------------------ Adj R-squared = 0.0025 Total | 113.653234 1524 .074575613 Root MSE = .27275 ------------------------------------------------------------------------------ ln_cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- inv_m120k | -.190606 .0872939 -2.18 0.029 -.361835 -.0193771 _cons | 4.287814 .0294011 145.84 0.000 4.230143 4.345485 ------------------------------------------------------------------------------ . capture drop stdres . predict stdres, rstandar (49 missing values generated) . twoway (scatter stdres milk120k) (lowess stdres milk120k) . end of do-file . do "C:\Users\javier\AppData\Local\Temp\STD01000000.tmp" . fp , scale center replace: reg ln_cf (fitting 44 models) (....10%....20%....30%....40%....50%....60%....70%....80%....90%....100%) Fractional polynomial comparisons: ------------------------------------------------------------------------------- milk120k | df Deviance Res. s.d. Dev. dif. P(*) Powers -------------+----------------------------------------------------------------- omitted | 0 367.951 0.273 10.533 0.033 linear | 1 366.159 0.273 8.741 0.033 1 m = 1 | 2 361.396 0.273 3.978 0.138 -2 m = 2 | 4 357.418 0.272 0.000 -- -2 3 ------------------------------------------------------------------------------- (*) P = sig. level of model with m = 2 based on F with 1520 denominator dof. Source | SS df MS Number of obs = 1525 -------------+------------------------------ F( 2, 1522) = 5.27 Model | .782289941 2 .391144971 Prob > F = 0.0052 Residual | 112.870944 1522 .074159622 R-squared = 0.0069 -------------+------------------------------ Adj R-squared = 0.0056 Total | 113.653234 1524 .074575613 Root MSE = .27232 ------------------------------------------------------------------------------ ln_cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- milk120k_1 | -.5301266 .1654544 -3.20 0.001 -.8546694 -.2055839 milk120k_2 | -.0008516 .0004271 -1.99 0.046 -.0016894 -.0000138 _cons | 4.238434 .008295 510.96 0.000 4.222163 4.254704 ------------------------------------------------------------------------------ . fp plot, r(none) ytitle(Predicted Ln(cf)) . end of do-file . do "C:\Users\javier\AppData\Local\Temp\STD01000000.tmp" . reg ln_cf milk120k_1 milk120k_2 Source | SS df MS Number of obs = 1525 -------------+------------------------------ F( 2, 1522) = 5.27 Model | .782289941 2 .391144971 Prob > F = 0.0052 Residual | 112.870944 1522 .074159622 R-squared = 0.0069 -------------+------------------------------ Adj R-squared = 0.0056 Total | 113.653234 1524 .074575613 Root MSE = .27232 ------------------------------------------------------------------------------ ln_cf | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- milk120k_1 | -.5301266 .1654544 -3.20 0.001 -.8546694 -.2055839 milk120k_2 | -.0008516 .0004271 -1.99 0.046 -.0016894 -.0000138 _cons | 4.238434 .008295 510.96 0.000 4.222163 4.254704 ------------------------------------------------------------------------------ . capture drop stdres . predict stdres, rstandar (49 missing values generated) . twoway (scatter stdres milk120k) (lowess stdres milk120k) . end of do-file . exit, clear