Tipultech logo

Analysis of Frame Data

Sarah Ogilvie

Dependency of Observations

To avoid the statistical difficulties associated with dependent observations within dyads, all analyses were conducted at the dyad level. For each dyad, two measures of frame use were calculated. The first measure was calculated by adding the frequency of frame use for the two negotiators within each dyad (joint frame scores). This measure was used to analyse aggregated frame use within each negotiation. The second measure was determined by calculating the absolute difference between the scores of each negotiator (frame difference scores). Frame difference scores provided a measure of the discrepancy in frame use between the two members of the dyad and were used to observe relative frame shifts across time.

Length of Negotiations

The duration of the negotiations ranged from 10 to 55 minutes. The varying lengths of the negotiations were reflected in the range of negotiator utterances per negotiation and therefore in the number of frames that were coded (30 - 266 total utterances). The number of utterances differed substantially between conditions. Negotiators in the competitive condition expressed a total of 1878 utterances, compared with 1402 utterances in the cooperative condition and 1370 in the mixed condition. Because of these differences in length, all subsequent analyses were applied to the proportion of times each frame was used rather than on the raw frequencies. In using proportional data, the sum of the frame proportions sum to one. To analyse data of this kind using analysis of variance techniques, one or more of the proportions needed to be excluded from the analyses to eliminate the linear dependency within the set of variables (Harris, 1985). In the current study, there were eight frames of interest, with an additional other frame category occurring 10% of the time. Because this other frame category accounted for 10% of frame use, and was not of theoretical interest, this classification was omitted from subsequent analyses. Omitting this category yielded eight linearly independent frame scores.

Transformation of Frame Data

An examination of the distribution of the eight frame proportion scores across conditions suggested that, overall, they were substantially positively skewed. To reduce the skewness and to stabilise the variances of the frame proportions, all frames were arcsine transformed (2 x arcsin p; Winer, 1971). To conduct these transformations on the joint frame scores data, frequencies of frame use within dyads were converted to proportions and arcsine transformed. For frame difference scores, the applicant's and recruiter's frame proportions were individually arcsine transformed and the absolute difference was then computed and subtracted from the competitor's score. Arcsine transformation reduced skewness for all frame proportion data.

<< >>