Dear Statalist,
I am using using MANOVA to see whether mean yield changes with farming
system (bet subject factor) and if these changes are influenced by time
(interaction of farming sys and time). Based on the results (see below),
there is no significant difference for both variables. But I run an xtpcse
with these same variables, i.e. cat* (farming system), and the Tcat1&Tcat2
as the interaction variables. But the results say both have significant
effect. In addition to that, after xtpcse, I test whether cat1=cat2 and I
get high p value so as to accept the Ho. How can that be when the results
just say each category is significantly influencing yield.
I am wondering whether my mistake is in the program or in the
interpretation (or both?). I would highly appreciate your help.
Cris
RESULTS OF MANOVA
/*test of the between subject factor*/
. manova yield1 yield2 yield3 yield4 yield5 = farming system
Number of obs = 63
W = Wilks' lambda L = Lawley-Hotelling trace
P = Pillai's trace R = Roy's largest root
Source | Statistic df F(df1, df2) = F Prob>F
------------+--------------------------------------------------
farming system | W 0.9520 1 5.0 57.0 0.57 0.7193
e
| P 0.0480 5.0 57.0 0.57 0.7193
e
| L 0.0504 5.0 57.0 0.57 0.7193
e
| R 0.0504 5.0 57.0 0.57 0.7193
e
|--------------------------------------------------
Residual | 61
------------+--------------------------------------------------
Total | 62
---------------------------------------------------------------
e = exact, a = approximate, u = upper bound on F
. mat ymat = (1,0,0,0,-1\0,1,0,0,-1\0,0,1,0,-1\0,0,0,1,-1)
. mat list ymat
ymat[4,5]
c1 c2 c3 c4 c5
r1 1 0 0 0 -1
r2 0 1 0 0 -1
r3 0 0 1 0 -1
r4 0 0 0 1 -1
.
. /*test of the farming system*time interaction*/
. manovatest farming system, ytransform(ymat)
Transformations of the dependent variables
(1) yield1 - yield5
(2) yield2 - yield5
(3) yield3 - yield5
(4) yield4 - yield5
W = Wilks' lambda L = Lawley-Hotelling trace
P = Pillai's trace R = Roy's largest root
Source | Statistic df F(df1, df2) = F Prob>F
------------+--------------------------------------------------
label_group | W 0.9550 1 4.0 58.0 0.68 0.6066
e
| P 0.0450 4.0 58.0 0.68 0.6066
e
| L 0.0471 4.0 58.0 0.68 0.6066
e
| R 0.0471 4.0 58.0 0.68 0.6066
e
|--------------------------------------------------
Residual | 61
---------------------------------------------------------------
e = exact, a = approximate, u = upper bound on F
RESULTS OF XTPCSE
xtpcse yield cat* fert labor labor2 Tcat1 Tcat2, noconstant
Number of gaps in sample: 1
Linear regression, correlated panels corrected standard errors (PCSEs)
Group variable: convfarm2 Number of obs =
365
Time variable: year Number of groups =
73
Panels: correlated (unbalanced) Obs per group: min =
5
Autocorrelation: no autocorrelation avg =
5
Sigma computed by casewise selection max =
5
Estimated covariances = 2701 R-squared =
0.8716
Estimated autocorrelations = 0 Wald chi2(7) =
62402.29
Estimated coefficients = 7 Prob > chi2 =
0.0000
----------------------------------------------------------------------------
--
| Panel-corrected
| Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+--------------------------------------------------------------
--
cat1 | 318.1316 13.81461 23.03 0.000 291.0555
345.2078
cat2 | 295.4165 16.56159 17.84 0.000 262.9564
327.8766
fert | .0662363 .0228123 2.90 0.004 .021525
.1109475
labor | -1.932159 2.189823 -0.88 0.378 -6.224134
2.359816
labor2 | .1188553 .0816233 1.46 0.145 -.0411235
.278834
Tcat1 | 5.182032 2.757123 1.88 0.060 -.2218297
10.58589
Tcat2 | 8.036667 3.557229 2.26 0.024 1.064626
15.00871
----------------------------------------------------------------------------
--
Maria Cristina DM Carambas
Junior Researcher/PhD Student
Zentrum f�r Entwicklungsforschung (ZEF)
Walter-Flex Str. 3
Bonn 53113, Germany
Tel: (49)228-739052
Fax: (49) 228-731849
e-m:[email protected]
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
