Introduction

I am a postdoctoral research fellow in the Responsible Innovation FSP. My backgroud in social science. Before Data School, I thought I could write my own code, but that was not true. I rather used to copy codes from the web. I work on both qualitative and quantitative data. It is exciting to learn that data visualisation is possible not only from quantitative, but also from qualitative data.

My Project

In this project, I am studying how people’s perception influences their acceptance of technological interventions for a large-scale restoration. The data for this study comes from a public survey, conducted in 2018. The survey was designed to examine Australian public’s perception. There are a total of six predictors and one outcome variable. I employed the Structural Equation Modeling (SEM) to test the following hyptheses.

1. Predictor 1 and 6 predict Outcome.
2. Predictor 2, 3, 4, and 5 predict Predictor 6

Preliminary results

Figure 1 shows public’s response on eight different technological interventions. Apparently, Tech 5 is preferred, followed by Tech 8, Tech 6, Tech 3 and Tech 1. Tech 4 is the least preferred technology. The correlations between all the study variables are shown in Figure 2, all correlations were statistically significant at 95% confidence level. The relationships between Predictor 5 and Predictor 2, 3, 4 and 5, and Outcome and Predictor 1 and 5 are presented in Figure 3. Nearly all the relationships are linear. We ran an SEM, Table 1 shows the Beta value and pvalue. Predictor 2 and 4 are non-significant predictors of Predictor 5. Figure 4 shows the relationships between all the predictor variables and the outcome variable. The width of the strands indicates the beta coefficients. Light blue coloured strands represent non-signigicant relationships at 95% confidence level. Finally, the results suggest the hypothesis 1 can not be rejected.

Plots from R

Figure 1: Participants’ response on each technology

Figure 2: Correlations between the study variables

Figure 3: Relationships between dependent and independent variables

Table

Independent variables	Dependent variables	Beta	pvalue
Outcome	Predictor 6	0.2936675	0.0000000
Outcome	Predictor 1	0.4542194	0.0000000
Predictor 6	Predictor 5	0.3751524	0.0000000
Predictor 6	Predictor 1	0.1816208	0.0000000
Predictor 6	Predictor 2	0.0121496	0.6708499
Predictor 6	Predictor 3	0.1946816	0.0000000
Predictor 6	Predictor 4	0.0220048	0.3779884

Figure 4: Relationships between the predictors and the outcome variable

My Digital Toolbox

My original data sets were in .dta format. I used the readstata13 to import the data into R. All the other steps were performed within R. The tidyverse package was especially beneficial to tidy up my data. Additionally, I used the lavaan packaage to run SEM. For the correlation plot, I used the ggcorrplot package (Figure 2). The ggpubr packaged helped to combine the regression plots (Figure 3). Finally, to plot the relationships between the outcome and predicting variables, I used networkD3 package (Figure 4).

My time went …

I think running the SEM model and visualising the model using netwrodD3 was the most challenging part. It was the first time that I visualised an SEM model using this package.

Next steps

In relation to this project, the next step for me would be to develop it as a manuscript for a journal submission (fingers crossed). Regarding data analysis and visualisation, I will keep exploring and be open to challenge my comfort zone.

My Data School Experience

In a nutshell, it was an amazing experience. The efforts and dedications that our instructors and helpers have put were simply outstanding. I thoroughly enjoyed all the challenges posted to us on Microsoft Teams. I appreciate the work behind the screen to put things in place for us to use, e.g. notes, attendance, files, data, and images. Finally, I love working on data. I will be using my skills a lot in my project.