Skip to main content
. 2021 Nov 4;184:111136. doi: 10.1016/j.jss.2021.111136

Table 3.

Sampling methods applied to address the research questions.

Research Question Sampling methods
RQ1: What ratios of users are satisfied/dissatisfied (happy/ unhappy) with the apps? All reviews were analyzed to generate sentiment charts of all reviews of each app.

RQ2: What level of diversity/variability exists among different reviews and their informativeness? All reviews were analyzed to generate box plots of reviews text length. Then, we used “stratified” random sampling to choose a few reviews for discussions in paper text: choosing a few “long” reviews (in terms of text length) and a few “short” reviews to discuss the issues via examples.

RQ3: What are the key problems reported by users about the apps? All reviews were analyzed via word-cloud visualizations. Then, we used “stratified” random sampling to choose a few reviews for discussions: choosing a few reviews by random from each group of key topics (words) appearing the most in reviews.

RQ4: By looking at the “positive” reviews, what aspects have users liked about the apps? The set of ”positive” reviews for each app was derived using the AppBot tool. We also calculated the ratios of positive reviews among all reviews of an app, and report in the paper (Section 4.4). Then, we used random sampling to choose a few reviews for discussions in paper text.

RQ5: What feature requests have been submitted by users in their reviews? The set of ” feature request” reviews for each app was derived using the AppBot tool. Then, we used random sampling to choose a few reviews for discussions in paper text.

RQ6: When comparing the reviews of Android versus the iOS versions of a given app, what similarities and differences could be observed? All reviews were analyzed to generate charts. To compare the problems reported for each OS version, we used random sampling to choose a few reviews for discussions in paper text.

RQ7: Is there a correlation between the number of app downloads and the country’s population size? All reviews were analyzed to generate charts.

RQ8: Are there correlations between the number of reviews and the country’s population or the number of downloads? And also, what ratio of app users has provided reviews? All reviews were analyzed to generate charts.

RQ9: What insights can be observed from the trends of review volumes and their sentiments over time? All reviews were analyzed to generate charts.