manova presentation 1
TRANSCRIPT
Presenter: Salam Sawadogo
M98330059
MULTIVARIATE ANALYSIS
Use of Phytase treated Distiller’s Dried Grains with Solubles on Chinese Catfish (Silurus asotus) diet
INTRODUCTION
Distiller’s dried grains with solubles(DDGS) is the dry residues after removal of alcohol by distillation of certain grains(corn,sorghum…..)
It contains high nutrients profile (protein, fat , minerals)>>>>>used in fish feed
However,it has been showed to contain a high level of a particular antinutritional factor (phytic acid)(Nourreddini and Dang,2008)
Phytase is a commercially available enzyme capable of breaking down phytic acid >>>release nutrients for fish
Therefore, this study investigated the effect of phytase in Chinese catfish (silurus asotus) growth performance
To achieve this objective,a 10 weeks experiment was conducted and fish were fed phytase treated diet
C1=control, no phytase
S1= diet 1000 U/kg
S2=diet 5000 U/kg
Tp=control,no phytase
T1=diet 1000 U/kg
T2=diet 5000 U/kg
Post spray Pretreatment
Triplicate group of fish
Iw=initial weightFw=final weightWg=weight gainFi=feed intakeFcr=feed conversion ratioSgr=specific growth rateSurv=survival
Data were subjected to one way Manova and differences were found to be significant at 0.05 probability level
data exp2;input trt$ iw fw wg FI FCR SURV SGR; cards;
C1 24.6875 35.76 44.85063291 16.65303571 1.503999613 75 0.287357852
C1 25.1 32 27.49003984 9.791666667 1.419082126 100 0.188350459
C1 24.475 32.3 31.97139939 10.04791667 1.284078807 100 0.215142539
S1 25.2 35.69 41.62698413 14.71875 1.403122021 100 0.269903584
S1 25.275 40.55 60.43521266 19.45696429 1.273778349 87.5 0.366606597
S1 25.5 38.34 50.35294118 18.27357143 1.423175345 100 0.3162713
S5 25.375 36.57 44.1182266 1 4.49583333 1.294848891 100 0.283426624
S5 24.275 35.88 47.80638517 16.70567857 1.43952422 87.5 0.303023564
S5 24.85 37.26 49.93963783 16.57025 1.335233683 100 0.314136531
tp 24.6375 36.1 46.5246068 17.96208333 1.567030171 100 0.296268865
tp 24.65 35.91 45.70655849 10.29982143 0.914185334 87.5 0.291926965
tp 24.7 34.325 38.96761134 11.38089286 1.182430427 87.5 0.255202844
tP1 25.875 40.3 55.74879227 16.05428571 1.11294875 87.5 0.343615514
tP1 25.2125 40.75 61.62617749 16.11601191.037233268 75 0.372342325
tP1 25.6875 42.142 64.0597845 23.43886905 1.424391391 75 0.383932387
tP5 25.325 36.642 44.69045269 18.61625 1.644856422 75 0.286499778
tP5 24.5 37.441 52.82312925 11.25321429 0.869533622 87.5 0.328909086
tP5 24.85 38.56 55.17102616 16.61228571 1.211691153 87.5 0.340733272
;
proc glm;
class trt;
model iw fw wg fi fcr surv sgr= trt;
manova h=trt/ printh printe short;
run;
Use class statement>>>>independent variableSpecify model by listing outcome variables to the leftManova>>>>hypothesized effect of treatments represented as h
The SAS System 16:42 Friday, March 26, 2011 77
The GLM Procedure
Class Level Information
Class Levels Values
trt 6 C1 S1 S5 tP1 tP5 tp
Number of observations 18
In Manova, SAS provides univariate and multivariate output separately for each dependant variables
Dependent Variable: iw
Sum of
Source DF Squares Mean Square F Value Pr > F
Model 5 2.00694444 0.40138889 1.36 0.2397
Error 12 1.43447917 0.11953993
Corrected Total 17 3.44142361
R-Square Coeff Var Root MSE iw Mean
0.583173 1.382444 0.345745 25.00972
Univariate output for initial weight
Probability large(>0.05),no significant difference among fish at the beginning of experiment
Dependent Variable: fw
Sum of
Source DF Squares Mean Square F Value Pr > F
Model 5 102.3959149 20.4791830 9.06 0.0009
Error 12 27.1341717 2.2611810
Corrected Total 17 129.5300866
R-Square Coeff Var Root MSE fw Mean
0.790518 4.060889 1.503722 37.02939
Probability small, enzyme has a significant effect on the fish final body weight
Dependent Variable: wg
Sum of
Source DF Squares Mean Square F Value Pr > F
Model 5 1097.053599 219.410720 5.39 0.0079
Error 12 488.386159 40.698847
Corrected Total 17 1585.439758
R-Square Coeff Var Root MSE wg Mean
0.691955 13.29215 6.379565 47.99498
P value very smallWeight gain is affected by the enzyme treatment
Dependent Variable: FI
Sum of
Source DF Squares Mean Square F Value Pr > F
Model 5 89.0355431 17.8071086 1.47 0.2687
Error 12 144.9218100 12.0768175
Corrected Total 17 233.9573531
R-Square Coeff Var Root MSE FI Mean
0.380563 22.46496 3.475172 15.46930
Large probability>>>no significant difference among treatment in term of feed intake
Dependent Variable: FCR
Sum of
Source DF Squares Mean Square F Value Pr > F
Model 5 0.11819349 0.02363870 0.44 0.8151
Error 12 0.65040691 0.05420058
Corrected Total 17 0.76860040
R-Square Coeff Var Root MSE FCR Mean
0.153778 17.95363 0.232810 1.296730
F small and Large probability>>>no significant difference among treatments
Dependent Variable: SURV
Sum of
Source DF Squares Mean Square F Value Pr > F
Model 5 703.125000 140.625000 1.80 0.1873
Error 12 937.500000 78.125000
Corrected Total 17 1640.625000
R-Square Coeff Var Root MSE SURV Mean
0.428571 9.866606 8.838835 89.5833
Probability greater than 0.05>>>No significant at 5%
Dependent Variable: SGR
Sum of
Source DF Squares Mean Square F Value Pr > F
Model 5 0.03084126 0.00616825 5.32 0.0083
Error 12 0.01391321 0.00115943
Corrected Total 17 0.04475447
R-Square Coeff Var Root MSE SGR Mean
0.689121 11.25914 0.034050 0.302425
Probability less than 0.05>>>>showing a significant difference among fish in term of specific growth rate
E = Error SSCP Matrix
iw fw wg FI FCR SURV SGR
iw 1.434479 -0.16795 -9.1111 2.157587 0.335909 7.083333 -0.04861
fw -0.16795 27.13417 109.0890 38.32794 0.101864 -71.0773 0.581283
wg -9.1111 109.0890 4 88.3861 139.1041 -1.61323 -325.521 2.603061
FI 2.157587 38.32794 139.1041 144.9218 7.688004 -157.373 0.747302
FCR 0.33590 0.101864 -1.61323 7.688004 0.650406 -4.11113 -0.00806
SURV 7.08333 -71.0773 -325.521 -157.373 - 4.11113 937.500 -1.77000
SGR -0.04861 0.58128 2.603061 0.747302 -0.00806 -1.770 0 0.013913
Manova outputError sum of squares and cross product matrix is shown below
Partial Correlation Coefficients from the Error SSCP Matrix / Prob > |r|
DF = 12 iw fw wg FI FCR SURV SGR
iw 1.000000 -0.026921 -0.344228 0.149642 0.347763 0.193155 -0.344119
0.9304 0.2494 0.6256 0.2443 0.5272 0.2496
fw -0.026921 1.000000 0.947635 0.611210 0.024248 -0.445643 0.946054
0.9304 <.0001 0.0265 0.9373 0.1270 <.0001
wg -0.344228 0.947635 1.000000 0.522867 -0.090516 -0.481074 0.998595
0.2494 <.0001 0.0667 0.7687 0.0961 <.0001
FI 0.149642 0.611210 0.522867 1.000000 0.791870 -0.426953 0.526279
0.6256 0.0265 0.0667 0.0013 0.1457 0.0647
FCR 0.347763 0.024248 -0.090516 0.791870 1.000000 -0.166488 -0.084766
0.2443 0.9373 0.7687 0.0013 0.5867 0.7831
SURV 0.193155 -0.445643 -0.481074 -0.426953 -0.166488 1.000000 -0.490273
0.5272 0.1270 0.0961 0.1457 0.5867 0.0890
SGR -0.344119 0.946054 0.998595 0.526279 -0.084766 -0.490273 1.000000
0.2496 <.0001 <.0001 0.0647 0.7831 0.0890
Correlation between SGR and FW and WG (r=94.6% and 99.8% respectively)
Correlation FCR and FI(R=79.2%)>>> more feed consumed, better utilization
H = Type III SSCP Matrix for trt
iw fw wg FI FCR SURV SGR
iw 2.006944 12.67091 38.37684 11.89295 -0.13211 -17.1354 0.199336
fw 12.67091 102.3959 332.4754 90.72241 -2.00351 -155.471 1.755223
wg 38.37684 332.4754 1097.053 291.0444 -7.17902 -514.330 5.81308
FI 11.89295 90.72241 291.0444 89.03554 -1.00458 -84.7760 1.539916
FCR -0.13211 -2.00351 -7.17902 -1.00458 0.118193 6.93488 -0.03819
SURV -17.1354 -155.471 -514.330 -84.7760 6.934882 703.125 -2.65317
SGR 0.19933 1.755223 5.81308 1.539916 -0.03819 -2.65317 0.030841
Type III sum of square and cross product matrix from which will be calculated the characteristic root and Eigen vectors
Characteristic Roots and Vectors of: E Inverse * H, where
H = Type III SSCP Matrix for trt
E = Error SSCP Matrix
Characteristic Characteristic Vector V'EV=1
Root Percent iw fw wg FI FCR SURV SGR
12.245 75.17 -1.6263 14.2358 -3.0264 -1.0783 12.6044 -0.01397 -31.2387
3.1590 9.39 0.6774 -7.1808 1.3922 0.4224 -5.1685 0.01822 58.5429
0.5613 3.45 -1.8712 3.3847 -0.7659 -0.0391 1.0632 0.02857 -11.8126
0.2917 1.79 8.5300 -5.6100 1.8000 -0.2491 3.6740 -0.00155 -69.1995
0.0332 0.20 -8.2339 5.4316 -1.3697 -0.5329 7.0534 0.00978 38.5228
0.0000 0.00 -5.9758 4.4414 -1.6556 -0.4763 5.7535 -0.00403 128.486
0.0000 0.00 0.99 -0.8000 0.0844 0.3604 -3.5624 -0.01393 0.00000
Percents listed indicate the amount of variability in the treatment effect that every Eigen value and vector account for
The first row>>a high percent and FW, FCR, SGR have highest magnitudes indicate they are the most factor affected by the enzyme
MANOVA Test Criteria and F Approximations for the Hypothesis of No Overall trt Effect
H = Type III SSCP Matrix for trt
E = Error SSCP Matrix
S=5 M=0.5 N=2
Statistic Value F Value Num DF Den DF Pr > F
Wilks' Lambda 0.00871035 1.65 35 27.67 0.0888
Pillai's Trace 2.30166882 1.22 35 50 0.2572
Hotelling-Lawley Trace 16.29090053 2.47 35 8.1573 0.0878
Roy's Greatest Root 12.24546614 17.49 7 10 <.0001
NOTE: F Statistic for Roy's Greatest Root is an upper bound.
The independent variable(phytase) has no effect according to the first 3 multivariate test statistics but Roy’s greatest Root showed a highly significant effect