Data Processing and Analysis

In this study qualitative content analysis was used to analyze both the interviews and the open-response questions that were included in the questionnaires. One of the strengths of qualitative content analysis is its aim to understand the phenomenon by studying the participants’

subjective views, attitudes and perspectives, bringing detail and depth to the analysis.

Content analysis is a systematic method to reduce the amount of data, which follows a certain, iterative sequence of steps (Schreier 2014, 170–171). In data-driven content analysis the analysis process is not guided by any theory or theoretical framework. Rather, it derives from the data itself without predetermined analysis units, following the research problem. (Tuomi &

Sarajärvi 2018, 81–82.) Data-driven content analysis starts with reducing the data by coding, meaning that the parts of the text that are relevant to the research are fragmented or summarized.

The codes that represent the same theme, feature or concept are sorted into clusters, which are then given a label describing their content. These clusters are further pieced together and labelled until they form subcategories and categories. The aim of this process is to move from the original, raw data towards theoretical concepts and conclusions. (Tuomi & Sarajärvi 2018, 85; 91–94.)

For this study, the analysis process started by transcribing the teachers’ and the parents’

interviews. Regarding the parents’ interviews, only the English interpretation was transcribed, although at times it was needed to fill in the interpretation by listening to the original Spanish

version to fully grasp the meaning as occasionally some part of the speech was misinterpreted or missed altogether. The researcher’s Spanish was good enough for this task. The overall amount of transcribed text was 106 pages (Times New Roman, font size 12, spacing 1,5). At this stage, it was necessary to contemplate which parts of the data were relevant for this research and would best answer the research questions. For the analysis to be purely data-driven, withdrawing from the presuppositions set by the theoretical framework or interview questions was needed. The analysis process followed the iterative sequence of steps until a saturation point was reached and no more categories could be formed. The categories emerged with a relative ease, though some rearranging and pruning was needed to done even after the actual analysis process in order to keep the research focused.

TABLE 3. Example of the Analysis Process of One of the Teacher’s Interview Step Analysis Process Example

1 Meaning Unit When they are involved, the kids take more of a sense of importance of it. If they struggle, they still realize how important it is. Like, “Okay, I need to learn this, maybe my parents aren't helping me or they can't help me in this case, they don't speak English but they know that it's important, they're pushing me to do this.

And so, I need to find a way to figure it out or something, do my best. (T 4)

2 Condensation When parents are involved, the child sees studying as important and tries harder 3 Code Student Motivation and Responsibility

4 Category Effects of Parental Involvement in Children’s Learning

5 Umbrella Category IMPORTANCE OF INVOLVEMENT ACCORDING TO TEACHERS

Regarding the questionnaires, the analysis process of the open-ended responses was similar to analyzing the interviews although transcription was not needed. Instead, the parents’ written responses were translated from Spanish into English before analyzing.

At first, the close-ended questions were meant to be analyzed using either exploratory factor analysis or correlation coefficient, but the assumptions were not met; the properties of the data were not suitable and sample size was not large enough (Fabrigar 2012, 23–26; Eddington 2015, 32). Also, as the 5-point Likert-scale data cannot be reliably treated as continuous (Jamieson 2004; Wu & Leung 2017), which complicated the use of mentioned analysis methods. In the end, the close-response data was entered into SPSS and the frequencies of the responses examined. Frequencies represent the distribution of survey data and show how many times a certain value occurs in it (Lavrakas 2008, 293), which was a convenient way to address the data as they give an overview of the parents’ responses. The way the parents’ responses were

distributed among the Likert-scale was telling since it was possible to explore how the majority, or the minority, of the participants experienced their children’s early language education and working together with the teachers—and whether and how they hoped to develop the partnership. These frequencies were reported in percentages and used to support the results drawn from the interviews, giving an idea of how the subjective views of individual parents may emerge in a larger population.

7 RESULTS AND DISCUSSION

The results depict two perspectives of the holistic phenomena, partnership and parental involvement, demonstrating how the teachers and parents work together to find best ways to support and help the children in the context of bilingual education. The findings display parents and teachers’ process and effort to build a partnership, some of the affecting factors that may either hinder or encourage involvement, ways to support the parents and improve—springing from both parties’ needs and experiences.