The Roles of Academic Self-Efficacy and Intolerance of Uncertainty on Decisional Procrastination in University Students during the COVID-19 Pandemic

Round 1

Reviewer 1 Report

Thank you for having me as a reviewer of this interesting piece on the role of academic self-efficacy (SE) and intolerance of uncertainty. The authors have sampled about 300 Italian psychology students and perform regression analysis, investigating into the relations of SE, intolerance of uncertainty, and procrastination. They find up to moderate effect sizes and conclude that further studies have to be conducted to gain more insight on procrastination in university students.

While I do think that the authors discuss an interesting and important topic, I also found several sections worthwile to revise, relating mainly to methodology and statistical analysis. As a disclaimer I have to say that I am not completely sure whether all of these issues can be solved adequately. But for now I will lay out my questions and comments in more detail, and recommend re-submission after major revisions.

Introduction:

I liked how the authors organized this section, and I believe it should be quite easy to follow their arguments, even for readers not from the field. The research questions are derived nicely from the literature. However, please note my comments on the discussion in the paper, which also relate to its introduction. I would also recommend to have a native speaker go through the manuscript before re-submission.

Materials and methods, results:

After having read p.6, I was a bit surprised that the authors did not use the measures for which they reported internal consistency in further analysis. So if the various factors are used to assess ASES and IUS and relate them to the
procrastination scale, reliability and validity measures (with the latter missing so far) should be provided for each factor used in the regression models. Otherwise it's almost impossible to say whether the findings are due to an actual relation between the variables, which we assume, or measurement error could have influenced the results.

To better understand the relations within a scale, I would ask the authors to add a full correlation table. To me, this would help to straighten up the third section as well (tables 1-2 could be removed, perhaps tables 5-6, too, tables 3 and 4 could be merged by dropping the R and F tests, but including coefficients for all three models - please check with APA standards).
The correlations within a scale are also needed to judge if multicollinearity issues might have come up, which to me would not be surprising giving the high internal consistencies of the measures.

While these aspects have mainly been on the technical side, I would recommend that the authors discuss their conceptual ideas to statistical modelling in more detail. Here are some questions that I would ask the authors to address in a revised version of the paper:

* What was the rationale to include all SE measures at the same time,
but adding the IUS measures one by one to the models?
(* Why not using a single measure of SE? See above.)
* Were there any interactions between SE and intolerance of uncertainty?
* Why weren't any control variables included in the regression models
(e.g., gender, age, year)?
* Why wasn't latent variable modelling used, which from my understanding is the state-of-the-art approach to dealing with measurement error?

Discussion and conclusions:

It was difficult for me to follow these sections, as I would naturally organise them differently. From my point of view, the conclusions should be derived from the rest of the paper, and they include an interpretation of the findings.
From my reading of the paper, the authors discuss not so much their actual study or its results and limitations, but they bring up several new references that are not linked clearly enough to the present study. I would recommend to include all the relevant studies in the introduction and only refer to them in the discussion to explain the findings of the present study (e.g., another study might have come to different conclusions). So I suggest the following outline for this section (only after having mended the issues mentioned above, which are more important in my view):

* Summarise your findings and give an interpretation
* Relate this to relevant literature, e.g. by discussing effect sizes,
differences, similarities, and clearly discuss limitations of the present study
* Give an outlook on future research, perhaps not so much to your own work

Finally, I would like to stress again that the paper tackles a meaningful research question, particularly during covid times. So I encourage the authors to address the issues raised in this review, and I wish them the best of luck in continuing their work.

Author Response

First of all, thanks for suggestions provided by the two reviewers about statistical analysis and discussion mainly.

We answer for the entire paper indicating each point modified according to both reviewers:

• Introduction: in this section, we restructured some paragraphs improving the quality of the English form but maintaining the original structure of the paper as appreciated by the reviewers.
• Materials and methods: as suggested by the second reviewers, we defined the two main hypotheses of our study and articulated the relative discussion accordingly to them.

Sample: we inserted mean and standard deviation of participants.

About each measure, we calculated the missing values for reliability statistics and construct validity (convergent and discriminant), reported after describing each measure and included the values of Cronbach’s Alpha and split-half method with the correction of Spearman-Brown coefficient to test the internal consistency of all scales.

As suggested by the first reviewer, we eliminated table 1 and 2, 3 and 4 and inserted those with discriminant validity for self-efficacy and tolerance of uncertainty.

• Results: we reported the full correlation table to consider the relations between the analysed factors (table. 4), and explicit the associations discussed in the main text about factors of self-efficacy, decisional procrastination and dimensions of tolerance of uncertainty.

Finally, as indicated by the first reviewer, we eliminated one of the three models of linear regression because its incompletion leaving the two principal models with and without the dimensions of intolerance of uncertainty in relation to self-efficacy. Consequently, we reported in table 5 all the values of regression models, together with collinearity statistics as defined in the APA standard guidelines. About this last point result demonstrated that there is no multicollinearity effects between the used measures.

We decide to maintain the partial correlations between the variables as frequently suggested by other reviewers in previous papers.

• Discussion: considering the relevant suggestions provided by the reviewers about the missing explanation of our results, we deepened this part of discussion linking the obtained results with the initial hypotheses and addressing future direction of research in our academic context. The focus of explanation is referred to the contribution of self-image in the perceived self-efficacy rather than that of interpersonal climate and other-oriented problem solving.

We hope that these revisions will satisfy the requests of reviewers.

Reviewer 2 Report

First of all, I would like to thank the editor for the opportunity to review this paper. Although the Covid-19 theme should be projected to future situations of uncertainty, the article could be of interest to the scientific community. The work is good, although the statistical analyses could be expanded. Even so, the limitation is understandable due to the characteristics of the sample.

The theoretical justification is adequate. Perhaps, there is an excess of quotations that makes it difficult to interpret a key idea. However, if the authors consider it so, it should not be a negative fact. Perhaps, mention should be made of gender differences in coping with the pandemic situation. For example: https://doi.org/10.1080/21642850.2022.2158831

Methods

·       Please make clear the objective of the study and, if you wish to mention any hypothesis, specify it clearly.

·        Add the average age of the population.

·       Is the internal consistency of the scales used for your study or in the original scale? It would be highly recommended to add both values.

Discussion and Conclusions. When we add quotes in conclusions, the main message seems to be distorted. Please discuss these ideas in the discussion and synthesize very well the finding and contribution of your study in conclusions.

Author Response

First of all, thanks again for the further revisions provided by one of the reviewers with regard to the methodology.

We answer for each question asked by the reviewer as follow:

-  * What was the rationale to include all SE measures at the same time, but adding the IUS measures one by one to the models? (* Why not using a single measure of SE? See above.)

As reported in regression model, we computed three models with only components of self-efficacy (model 1), components of self-efficacy and two dimensions of intolerance of uncertainty (model 2), and self-efficacy, IUS and independent variables (age, gender, years of course degree) (model 3).

- * Were there any interactions between SE and intolerance of uncertainty?

These interactions between SE and IUS are reported in the table of bivariate correlations, even if this objective is not of interest of this study in direct way.

- * Why weren't any control variables included in the regression models (e.g., gender, age, year)?

See the first comment.

- * Why wasn't latent variable modelling used, which from my understanding is the state-of-the-art approach to dealing with measurement error?

We used the bootstrap method to verify the stability of models controlling for measurement error (see table 5-6).

We hope that these revisions will satisfy the requests of the reviewer.

Round 2

Reviewer 1 Report

Thank you once again for having me as a reviewer of this very interesting paper. I acknowledge that the authors have apparently put a lot of effort into the revision and I feel that the paper has severely improved in terms of clarity and structure. The manuscript is organized very nicely and it is much more easy to follow the authors' argumentation in my opinion.

However, what remains untouched are the open questions with regard to methodology that I raised in my first review (copied and pasted below). Therefore for now, I remain with my original decision even though the paper has clearly improved. I believe that it is important to address these questions to make sense of the conclusions that the authors derive from their analysis.

---

I would recommend that the authors discuss their conceptual ideas to statistical modelling in more detail. Here are some questions that I would ask the authors to address in a revised version of the paper:

* What was the rationale to include all SE measures at the same time,
but adding the IUS measures one by one to the models?
(* Why not using a single measure of SE? See above.)
* Were there any interactions between SE and intolerance of uncertainty?
* Why weren't any control variables included in the regression models
(e.g., gender, age, year)?
* Why wasn't latent variable modelling used, which from my understanding is the state-of-the-art approach to dealing with measurement error?

Author Response

First of all, thanks again for the further revisions provided by one of the reviewers with regard to the methodology.

We answer for each question asked by the reviewer as follow:

-  * What was the rationale to include all SE measures at the same time, but adding the IUS measures one by one to the models? (* Why not using a single measure of SE? See above.)

As reported in regression model, we computed three models with only components of self-efficacy (model 1), components of self-efficacy and two dimensions of intolerance of uncertainty (model 2), and self-efficacy, IUS and independent variables (age, gender, years of course degree) (model 3).

- * Were there any interactions between SE and intolerance of uncertainty?

These interactions between SE and IUS are reported in the table of bivariate correlations, even if this objective is not of interest of this study in direct way.

- * Why weren't any control variables included in the regression models (e.g., gender, age, year)?

See the first comment.

- * Why wasn't latent variable modelling used, which from my understanding is the state-of-the-art approach to dealing with measurement error?

We used the bootstrap method to verify the stability of models controlling for measurement error (see table 5-6).

We hope that these revisions will satisfy the requests of the reviewer.