SHSM024 / SHSM030 RESEARCH METHODS AND ANALYTICAL PROCEDURES 2021

QUESTION 1:  RESEARCH DESIGN, METHODS AND ETHICS  (15 marks)

1.1      What is the purpose of obtaining informed consent in research? (3 marks)

1.2      When conducting a placebo controlled supplementation study, what is the purpose of blinding the researcher to the participant group? (2 marks)

Questions 1.3 through to 1.5 refer to information provided in the ‘Scenario’ below, please read the scenario before answering these questions

1.3      Did the researcher follow the correct protocol for obtaining consent? Explain your answer. (3 marks)

1.4      Is the design of this study ethical? Explain your answer. (4 marks)

1.5      Do you agree with the conclusions of the researcher? Explain your answer. (3 marks)

Scenario:

A researcher wishes to examine the effects of a physical activity intervention on obesity in teenagers. She intends to monitor the physical activity levels of participants in the study and then assess whether there is a relationship between levels of physical activity and levels of obesity. After this she will assign participants to either a physical activity group or a normal activity group. She obtains ethical approval from her university before starting the study and recruits   via  advertisements  on  Facebook.   When   she   receives  a   response   to   her advertisement she asks the parents of the respondent to consent for their child to participate in the study after which she sends them the full study details. Once consent is received she sends a diary to the parents and asks them to fill out the activity of their child at the end of each week and send it back to her. She monitors activity levels of participants for two months. For 10% of participants she has missing diary sheets so she completes the activity profile for those weeks based on the activity levels in the other weeks. Her sample size calculation stated that she should recruit 60 participants for her study, however, after 30 participants she observes a significant relationship between physical activity levels and obesity so she decides to stop recruiting participants at this stage. Once she has finished the monitoring stage she assigns participants to the physical activity intervention group or the normal activity group. In order to ensure that participants benefit from the study she assigns participants with the highest levels of obesity to the physical activity group. At the end of the intervention she finds that the total weight loss is higherin the physical activity group. Based on this she concludes that her intervention was effective in promoting weight loss.

QUESTION 2:  SYSTEMATIC REVIEWS AND META-ANALYSIS  (10 marks)

The Forest plot below reports the findings of a very recent ‘systematic review with meta-analysis’ on studies that had tested whether ‘creatine supplementation during resistance training’ reduced body fat% more than ‘placebo supplementation during resistance training’. The units of the outcome is ‘mean body fat% difference between the two groups (creatine and placebo).

2.1      Describe, in general terms, what the two main aims of any systematic review are (2 marks)

2.2      Based on the level of variation between the 23 study findings, indicate whether it was appropriate for the authors of the systematic review to produce a single summary estimate of all 23 findings (support your answer with relevant values and terms). (3 marks)

2.3      Summarise the  results and conclusion of the ‘quantitative synthesis’  (Forest plot)  that  the  authors  performed  as  part  of  their  systematic  review,  then compare them to the results and conclusion that would have been produced if a ‘qualitative synthesis’ approach been used to summarise the same 23 study findings (support your answer with relevant values). (5 marks)

QUESTION 3:  ANALYSIS OF VARIANCE [ANOVA]  (5 marks)

Use your knowledge of ‘one-way repeated measures ANOVA’ to answer part 3.1, and use the SPSS ‘two-way independent samples ANOVA bar chart’ below to answer part 3.2

3.1       In  relation  to  ‘one-way  repeated  measures  ANOVA’,   describe  what   the assumption of ‘sphericity’ is (2 marks)

3.2      The ‘two-way independent samples ANOVA’ bar chart below reports the pre- match  cognitive  anxiety  levels  (mean  &  95%CI)  of  footballers  grouped  by competition level (amateur and professional footballers) and by gender (female and male). Interpret the findings represented in this bar chart in lay terms – i.e. without using statistical terms. (3 marks)

QUESTION 4:  SAMPLE SIZE AND STATISTICAL POWER  (10 marks)

A researcher is designing a study to test the effectiveness of a 6-week ‘shoulder strengthening’ intervention on ‘serve speed’ in semi-professional male tennis players.

She has estimated that the ‘smallest meaningful’ effect that she wants her intervention to detect is a 5km/h (Cohen’s D = 0.5) higher serve speed - compared to the control group. Using this estimate of D=0.5, she calculates the sample size her study requires in order to achieve 80% statistical power at the 5% level of statistical significance.

Use the ‘G*Power’ output (below) to answer part 4.1 and use your knowledge of ‘sample size and statistical power’ to answer parts 4.2 through to 4.4.

4.1      According to the G*Power output (below), how many participants, in total, are required to provide complete data for this study (1 mark)

4.2      Explain why it would be unethical if the researcher decided to recruit a sample size that was 50% less than the one calculated by G*Power for D=0.5. (3 marks)

4.3       If the researcher decided that even a  1km/h  (Cohen’s  D=0.1)  higher serve speed was  important,  state whether the  re-calculated  sample  size  for  this smaller effect would be higher or lower than the sample size calculated for a 5km/h difference (D=0.5). (2 marks)

4.4      Explain  what  ‘80%  power’  (and  the  remaining  20%)  means,  in  terms  of probability. (4 marks)

QUESTION 5:  STEPWISE LINEAR REGRESSION  (25 marks)

With karate being included as a new Olympic sport for the 2021 (postponed from 2020) summer Olympics in Tokyo, Team GB wanted to identify young female karate students (aged 14-16y) who had the potential to represent GB at the 2024 Olympic Games karate event. Team GB designed a batch of four tests to objectively assess four key aspects of karate ability. These tests were ‘Kicking accuracy’ (%: high is good)’, ‘10 Push-ups time (seconds: low is good)’, ‘30 Punches time (seconds: low is good)’ and ‘Flexibility: splits score (%: high is good)’ .

In order to determine whether these tests were indeed predictive of karate ability, a random sample of 25 young female karate students (at black belt level) who had carried out all four tests, were assessed for their karate ability (%: high is good) during several sparring sessions by Team GB top karate coach/instructor. The five variables were used in a Stepwise linear regression analysis (via forward selection) with ‘Karate Ability rating score’ as the dependent variable, and the four tests as the independent variables.

Use your knowledge of Stepwise linear regression methods to answer part 5.1, and use the SPSS Stepwise linear regression output below to answer parts 5.2 through to 5.5:

5.1      Describe  the  processes  involved  in  Stepwise  linear  regression  (via  forward selection) and explain why it is relevant for this study. (6 marks)

5.2      Identify and interpret the following three (most common) tests of assumptions for multiple linear regression analysis: No multi-collinearity of the independent variables, Normality of the residuals, Homoscedasticity of residuals (6 marks)

5.3      Describe  how  the  regression  model  improved  at  each  ‘step’  of  the  model building  process  (support your answer with  relevant  values and terms).  (5 marks)

5.4      Interpret the ‘unstandardized B coefficients’ for each statistically significant IV in the final model (4 marks)

5.5      Based  on  the  findings  of this study,  make a  recommendation to Team GB regarding the tests they should  use and  make a  recommendation to them regarding further research. (4 marks)

QUESTION 6:  MODERATED HIERARCHICAL LINEAR REGRESSION (15 marks)

A PhD student was given access to three data variables that were collected as part of a large cross-sectional study of n~500 elderly (aged 65-100y) women in the UK. The three variables were: ‘Frailty score’ (continuous dependent variable: %: high % = more frail),  ‘Age’  (continuous  independent  variable),  time  spent  walking  (dichotomous independent variable: 1 = Walks>1hr/wk, 0 = Walks≤1hr/wk).  It has been shown many times that frailty increases with aging, so the PhD student wanted to examine whether ‘time spent walking’ moderated this age-related increase in frailty.  In order to examine ‘moderation’, the student entered an ‘Age x Walks>1hr/wk’ interaction term in to a ‘hierarchical linear regression’ model.

Use your knowledge of ‘moderated  hierarchical regression’ and the SPSS output below to answer parts 6.1 through to 6.3, and use your knowledge of study designs to answer part 6.4.

6.1      Describe the model building process involved in ‘hierarchical linear regression’ and explain why it is relevant for this study. (5 marks)

6.2      Using the information in Model 3 of the Regression table, predict the Frailty score(%) for an 80y old woman who walks for more than 1hr/wk. (3 marks)

6.3      Using the information in Model 3 of the Regression table, predict the Frailty score(%) for an 80y old woman who walks for less than 1hr/wk. (3 marks)

6.4     The  PhD  student  concluded  that  “walking  for  more than  1hr/wk  leads  to a decrease  in  the  Frailty  score  of  elderly  women”.  However,  the  student’s supervisor pointed out to him that they were unable to say “leads to” because their study was cross-sectional and that such study designs cannot determine causality. Provide a concluding sentence that is relevant to this cross-sectional study design,  and then  recommend  a  study  design  that  would  be  able  to determine causality – that walking >1hr/wk leads to a decrease in Frailty score.

(4 marks)

QUESTION 7: LOGISTIC REGRESSION  (15 marks)

An epidemiology researcher analysed data collected from a ‘car ownership & physical activity’ survey of n~2500 adults living in rural villages in Devon, UK. The dependent variable  was  analysed  as  a  dichotomous  variable  (meets  the  recommended 150min/wk of moderate-and-vigorous intensity physical activity [MVPA]: Yes=1 or No=0)   using   cross-tabulations   and    logistic   regression.   The   two   categorical independent variables were ‘no. of cars in the household’ (0 cars, 1 car, ≥2 cars) and ‘Age’ (18-34y, 35-49y, 50-64y, ≥65y)

7.1      Using the  percentages  reported  in the 3x2 cross-tabulation below, calculate the Odds Ratio for meeting the ‘MVPA≥150min/wk’ recommendation when there are ‘≥2 cars’ in the household, compared to ‘0 cars’. (note that ‘Age’ is not factored into this calculation) (4 marks)

7.2      Using  the  information  in  the  Logistic  Regression  table  below,  report  and interpret the Odds Ratio of ‘meeting the MVPA≥150min/wk recommendation’ for ‘≥2 cars’ versus the reference group of ‘0 cars’ (reference group is not reported) when Age is factored into the model. (4 marks)

7.3      Using  the  information  in  the  Logistic  Regression  table  below,  report  and interpret the Odds Ratio of ‘meeting the MVPA≥150min/wk recommendation’ for ‘Age65y’ versus the reference group of ‘Age18-34y’ (reference group is not reported) when ‘no. of cars’ is factored into the model.  (4 marks)

7.4       Explain, in general terms, what the ‘Cox & Snell R square’ value is used to assess and then interpret the value reported in the  Model Summary table below. (3 marks)