Teste de comparação de médias
As médias podem ser obtidas da seguinte maneira:
> tapply(morango$resp, morango$trat, mean)
A C G L
0.1711125 0.2212875 0.4393375 0.1794000
O teste de Tukey é realizado da seguinte forma:
> morango.tu <- TukeyHSD(morango.av)
> morango.tu
Tukey multiple comparisons of means
95% family-wise confidence level
Fit: aov(formula = resp ~ trat, data = morango)
$trat
diff lwr upr p adj
C-A 0.0501750 -0.01692822 0.11727822 0.1973032
G-A 0.2682250 0.20112178 0.33532822 0.0000000
L-A 0.0082875 -0.05881572 0.07539072 0.9865109
G-C 0.2180500 0.15094678 0.28515322 0.0000000
L-C -0.0418875 -0.10899072 0.02521572 0.3404304
L-G -0.2599375 -0.32704072 -0.19283428 0.0000000
Para visualização de forma gráfica, basta apenas fazer um plot do objeto morango.tu:
> plot(morango.tu)
No pacote agricolae há uma função que representa o resultado do teste de Tukey com as letras indicando os grupos.
> require(agricolae)
> df <- df.residual(morango.av)
> MSerror <- deviance(morango.av)/df
> tukey.tu <- HSD.test(morango$resp, morango$trat, df, MSerror,
+ group = TRUE, main = "Comparação de médias - morango")
Study: Comparação de médias - morango
HSD Test for morango$resp
......
Alpha 0.050000000
Error Degrees of Freedom 28.000000000
Error Mean Square 0.002416141
Critical Value of Studentized Range 3.861243662
Treatment Means
morango.trat morango.resp std.err replication
1 A 0.1711125 0.01162003 8
2 C 0.2212875 0.02418321 8
3 G 0.4393375 0.01797667 8
4 L 0.1794000 0.01284746 8
Honestly Significant Difference 0.06710322
Means with the same letter are not significantly different.
Groups, Treatments and means
a G 0.4393375
b C 0.2212875
b L 0.1794
b A 0.1711125