Chapter Six Methods and procedures
7. Analysis 1 Data entry
7.2. The creation of tables and matrices for the analysis of group data
1. Contingency tables showing mothers' moves and the acts within them followed by children's moves and acts within those, were created for each mother-child pair using Minitab Version 10. Tables were also created for child moves and acts followed by mother moves and acts. These tables were then copied into matrices, which permitted sequential analysis of the data.
2. The large matrices containing acts within moves (mother-child and child-mother) were collapsed to show the fi'equencies of moves produced by mothers followed by moves produced by children (from the mother-child matrices), and of moves produced by children followed by those produced by mothers (from the child-mother matrices). A similar conversion was also made to create act-act frequency matrices for mother-child and child-mother data. The formation of the matrices is represented diagrammatically in Appendix 9, also shown are the macros 'mactr%f.mtb' and 'cr%f.mtb' used for the creation of the matrices.
3. Data appearing in rows in the above matrices (described in 7.2.2.) were summed to calculate the move and act frequencies for mothers (matrices arising fi'om
'mactr%f.mtb') and children (matrices arising from 'cr%f.mtb').
4. The frequency data for moves and acts produced by individual mothers and children were transformed into proportions. That is, the number of times a move or an act was used by an individual mother or child divided by the total number of moves or acts produced by that subject. These transformations were also calculated using the macros 'mactr%f.mtb' and 'cr%f.mtb'. (See Appendix 9.)
5. Frequencies of individual moves and acts produced by mothers and children in each group (four groups: mothers Group 1 and 2, children Group 1 and 2) were summed and divided by the total number of moves or acts produced by the group, to give the mean proportions of moves and acts produced by each group. This calculation was also performed using the proportions of moves and acts produced by individual subjects, in 7.2.4. above. (See Appendix 9 for a diagram of the formation of tables for these calculations.) Results of these calculations showed that there was very little difference between the average proportions of moves and acts created from individuals' frequency data and those created from their proportional data. This indicates that frequencies of acts and moves were quite similar within groups. It was decided that the best way of looking for most interaction trends was through proportional data, which minimises differences which arise from one subject producing more communication behaviours than most of the other subjects. Results of these calculations, for each of the four groups of subjects, are found in Appendix 10.
6. The proportions of moves and acts produced by both groups of mothers and children were examined from Appendix 10. Further reductions in the acts and moves to be analysed were made by amalgamating some acts and omitting others. Acts were amalgamated if:
- they had a very similar function, for example the types of request for clarification produced by children,
- and they occurred infi-equently (less than 4%) in the data,
- and they were produced in similar proportions across the two groups of mothers and children.
Acts and moves were omitted if the reliability study had not indicated a high consensus in their coding. In some instances the dummy code (100), which was added to the original data to create a strict mother-child alternation, was also omitted. The dummy code was retained in child data in mother-child act and move analysis, and in mother data in child- mother act and move analysis. Retaining the code in mother-child data allowed calculations to be made of how frequently mothers produced an act or move and then immediately produced another act or move themselves, or how often the children did so fi-om the child-mother data. The 100 code was omitted in mother data in mother-child act and move analysis because it was always followed by a child initiation, and in child data in child-mother analysis because it was always followed by an initiation by the mother. (See Appendix 10.) As sequential analysis was to be undertaken at the lag 1 level only, there was no way of observing which act preceded the dummy code, to which the act following the dummy code would be in relation.
The moves and acts retained for analysis for mothers and children are given in Table 19 and Table 20 below.
Table 19 Moves retained for the analysis o f mother and child data
Mother ChUd
Initiation (I) Initiation (I)
Response/Initiation (R/I) Response/Initiation (R/I)
Response (R) Response (R)
No Response + No Response Requested No Response
(NR) (NR)
Follow-up (F) No Response Requested (NRR)
(100 in child-mother analysis) (100 in mother-child analysis)
Table 20 Acts retained for the analysis of mother and child data
Mother ChUd
Request Joint Attention (plus string+RJA) (RJA)
RJA (plus string+RJA)
Request for information (plus string+RI) (RI)
RI (plus string+RI)
Request for object/action (ROA) ROA (plus string+ROA)
string+ROA PI
Provision of information (plus string +PI) (PI)
string +PI
Expression of self (ES) ES
Acknowledgement/Confirmation/Denial (AK+CD)
CD
Request for Clarification; Confirmation (RCC)
AK
Request for Clarification: Neutral (RCN) Unintelligible
Request for Clarification: Specific (RCS) RClar (RCC+RCC+RCS) Provision of Clarification (PClar) PClar
No Response (NR) NR (100+NRR omitted in mother-child
analysis)
(100+NRR omitted in mother-child analysis)
7. The probability of children producing each type of move following individual moves made by mothers was also calculated, using the reduced range of moves in Table 19 above. For example, the probability of children performing an initiating move following mothers' production of a response move (ie. P Child Move Y if Mother Move X). In this calculation the probability of mothers producing each type of move summed to 1. A similar calculation was also made of the probability of children producing each type of act following individual mother acts. The probabilities of mothers producing each type of move and act following children's moves and acts were also calculated.
8. Contingency tables containing child moves plotted against the modes used to express them were created from the original Minitab worksheets. These tables were copied into matrices. Each matrix was compressed to give information about the frequency of mode types used by individual children and the proportions these frequencies represented (using 'mode%.mtb', shown in Appendix 9).
9. A similar operation to that performed in 7.2.8. was carried out to give matrices showing to modes used by children to express individual act types.