Lecture Five 2014
Uploaded by
danLecture Five 2014
Uploaded by
danLecture 5a: Writing-Up Results
Lecture 5b: Online Survey Engines
Thinking about presenting our results
Using google docs (and survey monkey) for an online survey
Other online survey tools (thanks to Dr. Steve Bush)
References:
Book 10 of Survey Kit covers reporting on surveys
Other examples on UTS Online
The SPARKPlus question in first assignment...
Our Research Question
Psychological and social factors can influence the survival of
patients with serious diseases. One study, published in Public
Health Reports 1980, examined the relationship between the
survival of patients with coronary heart disease (CHD) and pet
ownership. Each of 92 patients was classified as having a pet or
not, and by whether they survived for one year.
The researchers thought that having a pet might improve survival.
Do you think there seems to be any evidence for this theory?
SPSS Output
Basic SPSS cross-tabulation...
We find that 71.8% of patients with no pet survived compared
to 94.3% for those with pets – BUT is it significant?
SPSS Output...
Need this to be less than 20% of cells – otherwise consider
collapsing categories
With a p-value of 0.003 we reject our null hypothesis: there is
a significant association between owning a pet and survival...
How should we report this?
Did you survive?
Died Survived Total (Counts)
Did you own a No Pet 28.2% 71.8% 100% (39)
pet? Owned a Pet 5.7% 94.3% 100% (53)
Overall 15.2% 84.8% 100% (92)
Source: blah blah
Significant association (p-value less than 0.5%) between ownership and survival.
Based on the crosstabulation we can see that the survival rate for pet owners is 94.3%
while it is only 71.8% for those with no pet. Based on a chi-squared test of association,
this differential in survival is highly significant.
We discuss the pattern in the data
o IF we have done a test we can claim a significant (or non-
significant) effect
o Add raw SPSS output to appendix...
Our Research Question
In her studies for her PhD, Holt examined the satiating properties
of various foods (see Holt, S., Brand Miller, J., Petocz, P. and
Farmakalidis, E. (1995) “A satiety index of common foods”,
European Journal of Clinical Nutrition, 49, 675-690). Volunteers
were fed iso-energetic portions (1000kJ) of various foods, and
their satiety was measured over the following two hours.
A boxplot show the distributions
across the three foods and the
potential for an effect.
We’d discuss this (without
claiming significance) and
also comment on the
variation – WHY?
SPSS Output
No evidence to reject the
null hypothesis of equal
variances...
Strong evidence (p-value =
0.001) to reject the null
hypothesis: at least one
group has a different mean
SPSS Output
This confirms that the difference is potato (with ice-cream and
croissant NOT significantly different).
How should we report this?
Confidence Interval
N Mean Std Error Lower Upper
potato 13 248.2 39.36 162.4 334.0
ice-cream 12 104.0 28.45 41.4 166.7
croissant 12 66.6 23.53 14.8 118.4
Total 37 142.5 22.15 97.6 187.5
From the table we can clearly see the mean satiety for potato is considerably higher
than for ice-cream and croissant. Using ANOVA (might add a footnote to confirm equal
variances), we reject that the population means are equal and accept that the foods do
have different mean satiety (p-value of 0.001). Pairwise comparisons with t-tests (using
Tukey’s adjustment for multiple comparisons) confirms that the mean for potato is
significantly higher than the other two, while there is no evidence of a difference in the
population means for ice-cream and croissant.
We can add raw SPSS output to the appendix to demonstrate
we have done the test correctly...
Online Survey Administration
One of the easiest ways to administer a survey is to create one
online
Assumes we have the mechanism to deliver our survey to a
suitable sample
o Lists of emails for a population
o Surveys of students / staff within the University
If our recruitment is less random we can use things like facebook
(and other ads) to promote our survey to interested respondents
May well be suitable for a pilot provided we can get coverage
of sub-groups of interest
o Can use quotas to adjust our achieved sample to look
more like the population of interest...
Google Docs
The product is now integrated into the Google Drive app
Has word processor, spreadsheet, presentation, drawing
applications
Also allows you to create forms, which can work as a
(simple) online questionnaire
To create a new survey:
Select Create > Form.
To add a heading:
Click in first text box (containing the text ‘Untitled form’)at the
top of the page and type the title.
To add an introduction:
Click in the second text box and type the introduction.
To add a question:
Update the question that is already showing when you open a
new form.
You can select different types of question
Select “Add item” at the bottom of the form to add the next
question.
To send:
Select “Send form” at the bottom of the screen...
To obtain responses:
Select “Choose response destination” at the top of the screen.
This creates a spreadsheet in google, which is easily moved into
excel for further manipulation...
Sorting Out Multiple Response Questions
This needs some manipulation in excel before moving to
SPSS...
1. Copy the column into notepad (or Word)
2. Replace ", " (comma followed by a space) by "," (comma without a space)
3. Copy back into Excel (in a new column)
4. After pasting, choose ‘Text to Columns’ under the Data tab
5. Choose “Delimited”
6. Check the Comma box, leave the rest blank.
7. Select Finish
8. In a blank cell, to the right of the new columns label, set up column labels
(in row 1) using the text exactly as in the option labels of the
questionnaire.
Sorting Out Multiple Response Questions
9. In the second row of the first newly created column, use the countif
function to count the number of times that category appears in each row
=countif($E2:$H2,I$1)
where E2:H2 contains the data pasted in steps 3 - 6 and I1 contains one of
the options. The absolute references allow you to copy across and down
without adjusting the formula.
10. To retrieve the “other” column (if one exists), use a formula like
=IF(COUNTIF(M$1:P$1,LOOKUP(REPT("Z",255),E2:H2))=0,
LOOKUP(REPT("Z",255),E2:H2),"")
(all one line) where M1:P1 contain the option labels and E2:H2 contains
the data that you pasted in steps 3-6.
11. Copy the new columns into SPSS.
An example posted on UTS Online...
Survey Monkey (www.surveymonkey.com)
Survey monkey is a specialist survey engine which has both free
and paid options. In the free option, you can create a basic
survey, with more advanced features (such as conditional
questions and downloading results into useful formats) reserved
for paid users. In the free version it is (more) difficult to export
your data to Excel/SPSS.
To create a new survey: Select “+ Create Survey”.
To add a heading: Enter the title after you create a new
survey.
To add an introduction: Use space under Add Page Title.
To add a question: Drag from Builder.
To send: Select the “Collect Responses” tab and follow the
instructions.
To obtain responses: Select “Analyze Results” tab.
Other Survey Providers
SoGoSurvey (www.sogosurvey.com)
Appears to allow users to download responses to excel. Setup
can be a bit lengthy, but the output is fairly easy to integrate
into SPSS from Excel. This is a good alternative to Google
Docs.
KwikSurveys (www.kwiksurveys.com)
Free, but does not appear to allow users to download results
Stellar Survey (stellarsurvey.com) / Fluid Surveys (fluidsurveys.com)
Need a paid account to download results
eSurvey Creator (www.esurveycreator.com)
A very good alternative to Google Docs, can download Excel
file in a sensible format. Setup is easier than SoGoSurvey.
Just to Finish...
Reporting on our survey is an important skill
More general than just this subject...
Think about what needs to be in the main report
SPSS output generally not included
o can be in an appendix (especially for your assignment)
BUT still think about what to include
If claiming a ‘significant’ difference / effect make sure that it’s
a statistical one...
Use simple charts to communicate patterns (avoid 3D)
...
Administering your survey online can be very effective
Google has a good (and free) tool for this but there are
others...
