Translate EXAMINING THE EXISTENCE OF RELATIONAL STRUCTURE
THIRD
STEP : EXAMINING THE EXISTENCE OF RELATIONAL STRUCTURE
Although indices based on individual codes offer
basic, and often relevant information. With this step we move increasingly
toward the analysis of interaction sequences, beginning with transactional
interchanges. The first step in analyzing these interchanges is to compose
transition tables based on the control direction of the behaviors observed. The
observations recorded in the mother-daughter interaction example can be
organized in terms of two types of transition tables, depending on wheter or
not the speaker order of contiguous message is taken into account. Table 3.4
represents all types of sequential interchange without considering the speaker
designation of interactors.
When
the speaking order of the interactors is of interest, the table entries are
organized by the antecedent (prior) or consequent (subsequent) position of each
interactor at the moment their behaviors are recorded. Therefore, to explore
the types of interchange between the mother and daughter based on speaker
order, two tables are composed, one with the mother in the antecedant position
as show in Table 3.5, and one with the daughter in this position as presented
in Table 3.6.
In
examining these tables, the first question concerns whether or not sequential
association exist in the table. For example, does the relational response of
the daughter tend to vary depending on the previous relational message of her
mother ? If we find that the behavior of relational control of the daughter is
associated with the control behavior of the mother, we can state that for this
interaction, relational structure exists. Once an association is established,
the spesific patterns of relational control that are responsible for or characteristic
of this structure can be examined. We are also able to compare these patterns
in different contexts in order to test spesific research hypotheses.
An
appropriate statistical method used to test the global existence of association
between-antecedant and consequent relational control behaviors in the
interaction is Pearson’s chi-square statistic. For the analysis of contingency
tables like those of our example, this statistical test indicates if a
significant association exists between the behaviors represented in the rows
and the behaviors represented in the columns. Thus, with the analysis based on
the whole table, this statistic is defined as.
Although
chi-square offers an estimate of the sequential association and therefore,
evidence of relational structure, a note of caution in needed in interpreting
the resulting degree of association, as this value increases with an increasing
sample size (total number of recorded codes). Another important aspect to keep
in mind is that the calculation of chi-square is based on an estimation of the
expected frequencies (fexp) of the contingency table studied. These
expected frequencies are estimated from the marginal frequencies of the table,
assuming no association between rows and columns. When the codes can be
repeated sequentially, the expected frequencies are defined as.
When
it is not possible for two similar codes to be recorded consecutively, a
procedure called “iterative proportional fitting” (Bakeman 7 Quera, 1995b) can
be used. However, the tables commonly produced with the relational control
coding procedures (as shown in Tables 3.4, 3.5, and 3.6) are based on
repeatable codes (a one-up code, for example, can follow another one-up code).
It
is also the case that the confidence in the chi-square value is not sufficient
when working with data tables that result in many, very low-expected
frequencies. For this reason, it is helpful when exploring relational structure
with this type of analysis, to use a program that readily provides the expected
frequencies or the proportion of expected frequencies of low value. Also note,
it is recommended that a similar statistic, the Likelihood-Ratio Chi-square, G2
(available in the GSEQ program) be utilized with tables containing more than
two dimensions and as well, with log-linear analysis.
To
return to our example, the results of analyzing the sequential association in
Tables 3.5 and 3.6, are given in Table 3.7. Here we observe that a significant
relational structure exists only in the interaction with the mother as
antecedent and the daughter as consequent. The p value is 328 for the daughter-mother interactions, but the p value approaches zero, 016, for the
mother-daughter interactions. These results indicate that the relational
structure is in one direction (i.e., the relational control behavior of the
daughter tends to change as a function of the type of control of the mother).
In the words, the findings suggest a unidirectional dependence. If an
association had been found with either the mother or the daughter as
antecedents. It would indicate a bidirectional dependence. For this example (as
shown in Table 3.7), there is a high percentage of very low-expected
frequencies which, other than for illustrative purposes, would caution against
making inferences about significance, as noted earlier, most relational control
analysis are based on many more observations than in the mother-daughter
example and thus, are typically not subject to the limitations of a small
database.
LANGKAH
KETIGA: KAJIAN KEBERADAAN STRUKTUR HUBUNGAN
Meskipun indeks
berdasarkan kode individu menawarkan informasi dasar dan relevan, dengan
langkah ini kita bergerak semakin ke arah analisis urutan interaksi, dimulai
dengan simpang susun transaksional. Langkah pertama dalam menganalisis simpang susun
ini adalah menyusun tabel transisi berdasarkan arah kontrol perilaku yang
diamati. Pengamatan tercatat dalam contoh interaksi ibu-anak dapat diatur dalam
dua jenis tabel transisi, tergantung pada keduanya, atau tidak urutan pembicara
pesan bersebelahan diperhitungkan. Tabel 3.4 merupakan semua jenis pertukaran
berurutan tanpa mempertimbangkan penunjukan ketua interactors.
Ketika
susunan berbicara tentang pembicara adalah kepentingan, pencatatan tabel oleh yang
di atas (sebelumnya) atau menyusul (berikutnya), setiap perilaku interactor
dicatat. Oleh karena itu, untuk mengeksplorasi jenis pertukaran antara ibu dan
anak berdasarkan urutan pembicara, dua meja tersusun, satu dengan ibu dalam
posisi atas sebagai pertunjukan pada Tabel 3.5, dan satu dengan anak perempuan
dalam posisi seperti disajikan pada Tabel 3.6 .
Table
3.4
Tabel
Kemungkinan untuk Kode Kontrol Hubungan
Antecedent
|
Consequent
|
|||
One-up (
![]() |
One-down (
![]() |
One-across (
![]() |
Totals
|
|
One-up (
![]()
One-down (
![]()
One-across (
![]()
Totals
|
15
7
2
24
|
7
8
5
20
|
3
4
6
13
|
25
19
13
57
|
Table
3.5
Tabel
Kemungkinan Pengendalian Hubungan Dengan Ibu
sebagai
anteseden dan Putri sebagai Pembicara Konsekuen
Antecedent
|
Consequent
|
|||
Mother
One-up (
![]() |
Mother
One-down (
![]() |
Mother
One-across (
![]() |
Totals
|
|
Mother One-up (
![]()
Mother One-down (
![]()
Mother One-across (
![]()
Totals
|
9
6
0
15
|
0
4
2
6
|
1
4
3
8
|
10
14
5
29
|
Table
3.6
Tabel
Kemungkinan Pengendalian Hubungan Dengan Putri
sebagai
anteseden dan Ibu sebagai Pembicara Konsekuen
Antecedent
|
Consequent
|
|||
Daughter
One-up (
![]() |
Daughter
One-down (
![]() |
Daughter
One-across (
![]() |
Totals
|
|
Daughter One-up (
![]()
Daughter One-down (
![]()
Daughter One-across (
![]()
Totals
|
6
1
2
9
|
7
4
3
14
|
2
0
3
5
|
15
5
8
28
|
Sebuah metode statistik
yang sesuai digunakan untuk menguji keberadaan global asosiasi antara perilaku-antecedant
dan konsekuen kontrol hubungan dalam interaksi adalah chi-kuadrat statistik
Pearson. Untuk analisis tabel kontingensi seperti dalam contoh, uji statistik ini
menunjukkan jika hubungan yang signifikan antara perilaku diwakili dalam baris
dan perilaku diwakili dalam kolom. Jadi, dengan analisis berdasarkan seluruh
tabel, statistik ini didefinisikan sebagai:
Meskipun chi-kuadrat
menawarkan perkiraan asosiasi berurutan dan karena itu, bukti struktur
relasional sebuah catatan diperlukan dalam menafsirkan di tingkat yang
dihasilkan dari asosiasi, karena nilai ini meningkat dengan ukuran sampel
meningkat (jumlah kode yang direkam). Aspek penting yang perlu diingat adalah
bahwa perhitungan chi-kuadrat didasarkan pada estimasi dari frekuensi yang
diharapkan (fexp) dari tabel kontingensi yang dipelajari. Frekuensi
ini diharapkan diperkirakan dari frekuensi marjinal meja, dengan asumsi tidak
ada hubungan antara baris dan kolom. Ketika kode dapat diulang berurutan, frekuensi
yang diharapkan didefinisikan sebagai:
Ketika tidak mungkin
untuk dua kode dapat mirip dengan direkam berurutan, prosedur yang disebut
"berulang pas proporsional" (Bakeman 7 Quera, 1995b) dapat digunakan.
Namun, tabel umumnya dihasilkan dengan kontrol relasional coding prosedur
(seperti yang ditunjukkan pada Tabel 3.4, 3.5, dan 3.6) berbasis kode berulang
(kode satu-up, misalnya, dapat mengikuti kode lain satu-up).
Tabel 3.7
Hasil Pengujian Struktur Relasional
Pearson’s
Chi-Square Result for the
Interaction
Daughter
![]() |
Pearson’s
Chi-Square Result for the Interaction Mother
![]() |
Pearson’s Chi-square = 4.621
Degrees of freedom = 4
Approximate p value = 0.328
(Expected frequencies < 5 = 88.9%)
|
Pearson’s Chi-square = 12.052
Degrees of freedom = 4
Approximate p value
= 0.016
(Expected frequencies < 5 = 77.8%)
|
Untuk kembali ke contoh
kita, hasil analisis asosiasi berurutan pada Tabel 3.5 dan 3.6, diberikan dalam
Tabel 3.7. Disini kita lihat bahwa struktur relasional yang signifikan hanya
dalam interaksi dengan ibu sebagai anteseden dan putri sebagai konsekuen. Nilai
p adalah 328 untuk putri-ibu interaksi, tetapi nilai p mendekati nol, 016,
untuk interaksi ibu-anak. Hasil ini menunjukkan bahwa struktur relasional dalam
satu arah (yaitu, perilaku kontrol relasional anak perempuan cenderung berubah
sebagai fungsi dari jenis kontrol dari ibu). Dalam kata-kata, temuan ini
menunjukkan ketergantungan yang searah. Jika asosiasi telah ditemukan dengan
baik ibu atau putri sebagai pendahulunya. Ini akan menunjukkan ketergantungan
dua arah. Untuk contoh ini (seperti terlihat pada Tabel 3.7), ada persentase
yang tinggi dari perkiraan frekuensi yang sangat rendah, selain untuk tujuan
ilustrasi, akan barhati-hati untuk tidak membuat kesimpulan tentang
signifikansi, seperti disebutkan sebelumnya, analisis kontrol yang paling
relasional didasarkan pada banyak pengamatan lebih dari pada contoh ibu-anak,
dan dengan demikian, biasanya tidak tunduk pada keterbatasan database kecil.
Comments
Post a Comment