Starting
your data analysis
Most students get quite excited when they finish
entering data and they have a data file to analyse. However, before
diving in to address all your research questions there are a few
things you need to do first. I have listed these below, along with
the related chapter in the SPSS Survival Manual.
| Check the characteristics of the subjects that
make up your sample. You will need this information for the
method section of your report. |
Chapter 6 |
| Check all the variables in your data file for
errors (particularly out-of-range values). |
Chapter 5 |
| Obtain descriptive statistics for each of the
variables you will be using in your study. These should include
means, standard deviations, kurtosis, skewness, and minimum
and maximum values. Check that these values are appropriate. |
Chapter 6 |
| Check
the distribution of scores on each of your variables—depending on the variable, you will need to
use histograms, boxplots, bar graphs or stem and leaf plots.
Look out for very skewed distributions or any unusual pattern
of scores. Also check for extreme outliers—these can affect
some analyses and may need to be recoded or removed. |
Chapter 5, 6, 7 |
| Perform the necessary data manipulation procedures
(e.g., recode, compute) to create any new variables you need.
This is important when creating total scores on a scale, or
collapsing down a variable into a smaller number of categories.
Afterwards, always run Frequencies on these new variables
to check that the procedure has been done correctly. |
Chapter 8 |
| Check the reliability of the scales you intend
using in your analyses. What are the Cronbach alpha values
for each scale? How do these results compare to those reported
in the literature? |
Chapter 9 |
| For your continuous variables, check the pattern
of intercorrelations. How strongly and in which direction are
your variables related? How does this compare with the results
reported in the literature? You may also need to obtain scatterplots
of the correlation between pairs of your major variables. These
are useful for checking for linear relationships between variables. |
Chapter 11 |
| When choosing which statistical technique to
use for your analysis, always check that you have the right
type of variables (categorical/continuous). Consider whether
a parametric or a non-parametric technique is the most appropriate. |
Chapter 10 |
| Check with your statistics books and the SPSS
Survival Manual to ensure you are not violating any of
the major assumptions for the analyses you intend to conduct.
This might involve checking that you have enough subjects
in your groups, that the variance for each group is similar,
or that the distribution of scores on your variables is not
too skewed. |
Parts Four and Five |
| Remember that SPSS will conduct the analyses
that you ask it to do, whether or not these analyses are appropriate.
The old saying 'Garbage in, garbage out' applies. It is up
to you to ensure that you understand what you are doing and
also what the output means. |
SPSS Survival Manual |
A few additional tips
1. Save
your output regularly so that if the computer crashes you have
not lost too much work. All output files should
be saved with a .spo extension onto your disk in the A:/ drive.
Give your output file a suitable name so you will able to identify
it later, for example 8aug96a.spo. Keep a list of your output files
with details of what is included. SPSS produces a lot of output
and it is very easy to get lost, so get organised—it will
save you a lot of time.
2. If you need to recode a variable, always create
a new variable. Keep the original variable so that if there are
any problems you have not lost the data.
3. If you create any new variables, always check
in your codebook that the name you intend to use has not already
been used. Otherwise you will lose all the original information.
Record the name and explanation of the new variable in your codebook.
Keep detailed notes of everything you do. This should include details
of cut-off points you use to recode variables, reasons for doing
things, reminders to yourself about how to do the analyses, problems
that might have occurred etc.
4. Finally,
make sure that when doing your analyses, you get up and stretch,
walk around etc., at least every hour.
SPSS for Windows can be addictive, a bit like eating peanuts—just
one more, and then, just one more … Plan what analyses you intend
to do, break your analyses into blocks, and give yourself time
to digest the output.
