Survey research and design in psychology/Assessment/Lab reports/4

Lab report 4: Multiple linear regression edit

Task: Conduct and report on a multiple linear regression which uses at least three independent variables to predict a dependent variable.

Marking criteria edit

In addition to the generic guidelines, this report should include:

  1. 5%. Title/Abstract: : Provide a succint overview of the study.
    1. Sumarise the study's purpose, method, findings, and implications in less than 150 words.
  2. 10%. Introduction:
    1. 5%. Background: Briefly introduce the topic and background literature about the research question and/or hypotheses.
    2. 5%. Research questions/Hypotheses: Propose logically-derived research question(s) and/or hypotheses (which is/are addressed in the Results). Research questions could be in form of "To what extent do A, B, and C predict the variance in Y?). Hypotheses could be in the form of "It is hypothesised that A, B, and C will each be positive linear predictors of Y").
  3. 10%. Method:
    1. Participants: N/A
    2. 10%. Materials: Explain the measures used for the variables in the current study, including summarising how any composite scores were derived and their internal consistency.
    3. Procedure: N/A
  4. 50%. Results: Describe and present the results of a multiple linear regression analysis involving three of more independent variables. Reporting should summarise any data screening and/or recoding/dummy coding (if not already covered in Materials), type of analysis, assumption testing, correlations between the variables, variance explained, and regression coefficients (including B, Beta, t, p, and sr2). No figures are required. The results should demonstrate a clear understanding of the nature of the linear relationships amongst the variables.
  5. 25%. Discussion: Explain what the analysis found out about the research question(s) and/or hypotheses e.g.,
    1. How much variance was explained in DV? Why?
    2. Which variable(s) contribute most/least to understanding the DV? Why?
    3. What were the study's strengths and weaknesses? What could be improved?
    4. What conclusions can be drawn and what are the implications?

FAQ edit

Should I use items or composite scores? edit
  1. Either
  2. If you're creating composite scores without having done an exploratory factor analysis first, then at least conduct reliability analysis and report this as part of the Method
  3. Well-derived composite scores (where available/possible) will generally provide for a more valid measure of fuzzy constructs than single-item scores, but this will depend on the specifics of the construct being measured.
Is my R2 big enough? edit
  1. Size of R2 doesn't matter in terms of meeting the marking criteria. In fact, R2 of 0 is acceptable. This is not an exercise in coming up with a large R2 , but rather one in which a theoretical model is tested to determine the extent to which a set of IVs predict a DV. Many models explain much less than 30% of variance. Although we might like to explain 30%, the reality is often much less. It might be just as important to know that some IVs don't predict a DV. This also gives you something to consider in the Discussion - how could the model and/or measures be improved?
Are my predictors significant enough or strong enough? edit
  1. The aim is conduct an MLR to examine how well some IVs predict a DV. Thus, it is not necessary to seek only significant predictors. It's more important to have a rationale for a model and then to test and interpret the results, regardless of their significance or size of effect.
What results should be presented in a table? edit
  1. The correlations and regression coefficients are best summarised in a table.
  2. For examples, see the MLR sample write-ups e.g., Table 12 from the sample report and Table 2 from the MLR report are good examples of tables which show the correlations and coefficients for a multiple linear regression analysis.

General feedback edit

Overall, performance was slightly better than on previous reports (67.9%).

  1. Word count
    1. Many assignments went over the maximum word count and thus received penalties
  2. Formatting
    1. Use page breaks rather than multiple line spaces to separate pages (better layout)
  3. APA style
    1. Formatting of citations was much better than previously. There is still some room for improvement - see previous feedback for suggestions.
    2. More reports used Australian spelling - see previous feedback for suggestions.
    3. Numbers were still often not formatted correctly - see previous feedback for suggestions.
    4. Spaces should be used either side of symbols which replace words (e.g., p < .05 instead of p<.05)
  4. Title
    1. Some titles lacked sufficient detail e.g., "factors predicting..." is vague - which factors?
  5. Abstract
    1. Generally, well done.
    2. The weakest aspect was too information about the design and not enough detail about the results e.g., for each predictor. Explain the significant, strength, and direction of the tested relationships.
    3. Avoid statistics in Abstract unless particularly pertinent.
  6. Introduction
    1. Literature reviews are becoming more succint and focused towards research questions - good to see.
    2. Research questions were generally clear and appropriate.
  7. Method - Materials
    1. Method sections are becoming more succint with important summary details
    2. Provide a citation to the survey and explain each of the items/composite scores and measurement scales etc. If using composite scores, report the number of items and internal consistencies.
  8. Results
    1. Generally, well done.
    2. Assumptions were well tested and described.
    3. Correlations sometimes were not included in a table and summarised.
    4. A table of regression coefficients was generally well presented, with effective description of the overall model.
    5. There was often a lack of sufficient explanation of the results for each of the predictors.
  9. Discussion
    1. Generally, well done.
    2. Stronger recommendations are those which are more specific and insightful; vague, unrelated recommendations are not convincing.
  10. References
    1. APA style in this section is improving.
    2. Do not include issue numbers for seriated journals (which is most of them).
    3. Journal article titles should be have the first letters capitalised.