Overview of Bivariate Measurement Data Report Writing

Key components of the statistical enquiry cycle for investigating bivariate measurement data:
• posing an appropriate relationship question using a given multivariate data set
• selecting and using appropriate displays
• identifying features in data
• finding an appropriate model
• describing the nature and strength of the relationship and relating this to the context
• using the model to make a prediction
• communicating findings in a conclusion.

Aiming for Excellence

Achievement Merit Excellence
Investigate bivariate Investigate bivariate measurement data, with Investigate bivariate measurement data, with statistical
measurement data involves justification involves linking components of insight involves integrating statistical and contextual
showing evidence of using each the statistical enquiry cycle to the context, knowledge throughout the statistical enquiry cycle, and
component of the statistical and referring to evidence such as statistics, may include reflecting about the process; considering
enquiry cycle. data values, trends, or features of visual other relevant variables; evaluating the adequacy of any
displays in support of statements made. models; or showing a deeper understanding of models.

Bivariate Data   
 Bivariate data compares two variables that are potentially connected e.g. ice cream sales and temperature
on that day
 In this assessment you will be given some raw data that you will be required to analyse by drawing a scatter
plot (using iNZight) and writing a report.

Overview of Time Series Report (Use headings 1-6 to organise your report)   
I notice / I wonder (do not include these notes in your final report) – use Scatter Plot Matrix in inZight
1. Introduction / Background
2. Identify features in the data (association, features, trend)
3. Select and justify an appropriate model (linear / non-linear)
4. Make a prediction in context with units and sensible rounding.
5. Examine other Models
6. Conclusion

1. Writing Introductions for Exemplar statements

  
Bivariate Measurement Data Italicised statements give alternative statements for describing the data
 Description of topic /  Measurements such as height, weight and lean body mass are useful ways of
variables from topic (one comparing athletes’ health and performance.
sentence).  Variables that might be likely to have a relationship are… because…
 This report will investigate if there a relationship between Haematocrit levels
Description and

and Red Blood Cell Count for athletes from the Australian Institute of Sport.
 Relationship Question (one  This report will investigate the nature of the relationship between Haematocrit
only). levels and Red Blood Cell Count for athletes from the AIS.
 This report will investigate if an athlete’s Haematocrit levels can be used to
predict their Red Blood Cell Count.
 Aim / Interest (Why worth  An understanding of the relationship between these variables might be useful
investigating? Questions?) to… because…
 Source  The source of this data is 120 athletes from the Australian Institute of Sport.
Data / Survey

 Definition and description

 The explanatory variable being investigated is…
of explanatory and
 The response variable being investigated is…
response variables.
 Important aspects of data  This data is likely to be valid as it is collected by doctors at the AIS.
collection details / validity.  This data was collected by xxx and therefore may not be a valid measure of …
 Link findings to what you
 In this context it seems likely that there would be a relationship between these
know / have researched
two variables as the Haematocrit level measures the percentage of red blood
Research

about the variables.

cells and the red blood cell gives a count of these same cells.
 Research suggests these
 One important aspect of this data is that all of the data points are for athletes
two variables may be
and so any relationships may not be applicable to non-athletes.
related because…
I understand… I need to work on…

2. Identify features Exemplar statements   

 Scatter Plot showing relationship between haematocrit level
and Red blood Cell Count for 120 Athletes at AIS.

Red Blood Cell Count

Inzight Graph: Scatter Graph (check
explanatory and response variables)
Graph

 Re-label graph axes if needed with

full title and units.

Haematocrit level

 What is nature of relationship?

 The scatter plot shows that as the Haematocrit level increases the Red
Association

 Inc / Inc or Inc / Dec

Blood Cell Count also increases.
 Justify by reference to visual
 The scatter plot shows that as x increases y decreases.
aspects.
 This is to be expected because…
(wider population – not just this
group).  This is to be expected as it is likely that the percentage of red blood cells
Context

 Name other variables that might may well impact on the number of such blood cells.
impact on the response variable  Other factors that may affect a person’s red blood count are… because…
and suggest how they might
impact. e.g. gender age.
 From the scatter plot it appears that there is a linear relationship
between Haematocrit levels and Red Blood Cell Count.
Trend

 Linear / non-linear?
 From the scatter plot it appears that there is a non-linear relationship
between x and y.
I understand… I need to work on…

3. Find a model Exemplar statements   

Scatter Plot showing relationship between haematocrit level
and Red blood Cell Count for 120 Athletes at AIS.
Red Blood Cell Count

Inzight Graph: Add to plot, select linear / non-

Graph

linear trend

Haematocrit level

 For the reasons given above a linear regression model has been
fitted to the data.
R Reason for linear / non-linear model
 For the reasons given above a non-linear regression model has
been fitted to the data.
Description of model  The linear model shows that Red Blood Cell Counts increase by
D
 Gradient statement if linear 0.1 for each increase of 1 in Haematocrit value.
 This model appears to be a good fit of the data throughout the
range of Haematocrit levels with all points aligning with the linear
trend.
 Discussion of fit throughout the range of x  The number of points above the trend line is also similar to the
values. number of points below.
 Look at how well the points align with the  However, there are no athletes with Haematocrit levels from 53
trend line for the range of x values. to 59 and so we are unable to describe the fit for this data range.
This means the model may not be as appropriate for assessing the
relationship between these variables when the Haematocrit levels
are over 52.
 Consider number of data points  This is a relatively high number of data points (120) which
enhances the reliability of the model.
 The relatively low number of data points means this model may
not be particularly reliable.
 Appro  This relationship is only statistical and does not imply that an
 Correlation / causation increase in Haematocrit level causes an increase in Red Blood Cell
Count.
 This relationship appears to be moderate-to-strong as there is
 Look at scatter some scatter along the trend line but it is not a large amount.
Strength of relationship

 Strong / moderate-to-strong / moderate /  The correlation coefficient is also relatively high at 0.93 indicating
weak-to-moderate / weak there is evidence of a fairly strong linear relationship between
 Look at amount of scatter about the Haematocrit values and Red Blood Cell Counts.
regression line  The scatter along the trend line is non-constant, with more scatter
 If linear then r / correlation coefficient after x. This suggests a stronger relationship for x = and a
 variation in scatter – constant / non- potentially weaker relationship for x =.
constant / fanning out  There is an increase in scatter after x = . This suggests the
relationship may not be as strong after this point.
 One unusual value is present with a Haematocrit level of 60 and
a Red Blood Cell Count over 6.5.
Unusual

 Visual description, numerical description,

 This value is along the same trend line as the rest of the data
discuss possible effect on model.
and so may inappropriately increase the strength of the
relationship.

 No groupings are apparent from the scatter plot.

Groupings /

 Visual description, numerical description,  Two groups are suggested in the scatter plot – the first with x < …
discuss possible reasons for differences. and the second with x > … One possible reason for these
differences may be…

I understand… I need to work on…

4. Make a prediction Exemplar statements   

Inzight Graph: as above
 Make a prediction for the response
variable using the equation of the trend
Prediction

line.  From this model I predict that the red blood count of a person with a
 Round answer sensibly and include units Haematocrit level of 50 will be 5.5. (¿ 0.11565∗50−0.26 ¿
if appropriate.
 Don’t relate to observed y-values.
 Given the moderate-to-strong relationship found in the data it is
 Justification regarding how accurate
likely that this prediction will be quite accurate.
prediction might be – reference to stat
 This prediction is likely to only be accurate for athletes as it is likely
evidence from analysis.
Justification

that they will have higher general Haematocrit and red blood cell
 Reflect on prediction by discussing their
levels that the rest of the population as they exercise more 1
relevance to wider population.
 Haematocrit is likely to be the best explanatory variable because…
 Justify choice of variables to use by giving
 Given the weak relationship found in the data this prediction is
reasons for using the selected one rather
unlikely to be particularly accurate and should only be taken as a
than others.
rough indication of y at point x.
I understand… I need to work on…

5. Further Considerations Exemplar statements   

Graph

1
http://www.livestrong.com/article/299082-the-effect-of-athletic-training-on-the-rbc-count/
  The data point at Haematocrit value 59 does not appear
to fit in with the rest of the data. This could be a valid
If unusual values:

 Comment on the effect any unusual values might

extreme value, or it also could be an error in
have on the model.
measurement.
 Justify why these values could be removed.
 For these reasons the model will be tested with and
 Extend the investigation by developing models with
without this data point to see the effect, if any, on the
data with and without the unusual values.
prediction made.
 As can be seen in the graph above…
Graph

 Comment on the effect the difference subsets might  One factor that may influence the relationship between
have on the model. Haematocrit levels and Red Blood count is the gender of
If subsets /
groups:

 Comment on the number of points now being the athlete.

investigated.  For these reasons the data will be split into these two
 Extend the investigation by developing models with groups and reanalysed.
data that has been separated into relevant subsets.  As can be seen in the graph above…
Re-predict

 Prediction made using alternative models

 Compare and contrast original prediction with updated
 Consider accuracy of these alternative models
(unusual value OR subset /group).
 Compare with original prediction

I understand… I need to work on…

6. Conclusion Exemplar statements   

 This report investigated whether a relationship exists between
 Give concise summary linked to original the Haematocrit levels and red blood cells for 120 athletes from
purpose of the investigation the AIS.
Summary

 Purpose of report  Analysis of the data showed a moderate-to-strong linear

 Brief description of model, including trend, relationship between Haematocrit levels and red blood cells.
strength and numeric.  This relationship showed that Red Blood Cell Counts increase by
0.1 for each increase by 1 in Haematocrit value.
 This model was then used to predict the Red Blood Cell Count
 What the model predicted
Prediction

for an athlete with a Haematocrit level of 50. The Red Blood

 Link to context
Count was predicted to be… This means…
 Accuracy of prediction
 This prediction is likely to be accurate because…
 Summary of investigation into other relevant
variable
Extension

 A further variable that was investigated was…

 Summary of what this means in context /
 Possible limitations of this model include…
research / future investigations
 These findings may be useful because…
 Usefulness / Limitations / Improvements /
Possible uses / Future Investigations

I understand… I need to work on…