4 Teste de Unidimensionalidade

Dentro do pacote ltm há uma função chamada unidimTest() que ode ser utilizada para testar a unidimensionalidade.

Utilize o arquivo soldi.

O arquivo pode ser obtido da seguinte maneira:

  > soldi<-read.spss('/home/adilson/Área de Trabalho/RTRI/dados/Soldi.sav',to.data.frame=T)
  > head(soldi)

    Id Afeti_1 Afeti_2 Afeti_3 Afeti_4 Afeti_5 Afeti_6 Inst_1
  1  1       0       0       0       0       0       0      1
  2  2       0       1       0       0       1       0      1
  3  3       0       1       0       0       0       0      0
  4  4       0       0       0       0       0       0      1
  5  5       1       1       0       0       0       1      0
  6  6       0       0       0       0       0       0      1
    Inst_2 Inst_3 Inst_4 Inst_5 Inst_6 Norm_1 Norm_2 Norm_3
  1      0      0      1      1      1      0      0      0
  2      1      0      1      1      1      1      1      1
  3      0      0      1      0      1      0      1      1
  4      0      1      1      0      0      0      1      1
  5      1      0      0      1      0      0      1      1
  6      1      1      1      1      1      0      1      1
    Norm_4 Norm_5 Norm_6
  1      0      0      0
  2      0      1      0
  3      0      0      0
  4      1      1      0
  5      1      1      1
  6      0      1      0

Considere inicialmente o modelo de Rasch.

  > soldi.ras<-rasch(soldi[,-1])

Realizando o teste:

  > unidimTest(soldi.rasch)
Alternative hypothesis: the second eigenvalue of the observed data is  
substantially larger than the second eigenvalue  
of data under the assumed IRT model  
 
Second eigenvalue in the observed data: 2.5036  
Average of second eigenvalues in Monte Carlo samples: 0.9551  
Monte Carlo samples: 100  
p-value: 0.0099

Experimente fazer o mesmo teste com os dados de LSAT do pacote ltm:

  > unidimTest(rasch(LSAT))