Practicals Flashcards
What are the two purposes of descriptive statistics and presentational techniques?
1) To summarise the data as clearly as possible in order for the viewer to digest the information with ease.
2) For a researcher to explore their own data.
How would you best present the summary of a single variable?
Box plot.
Especially informative if the data is not normally distributed.
Thick band shows the median, box shows quartiles and whiskers show the full range; some programmes also display outliers.
How would you best present the distribution of a single variable?
Bar chart: if the data is discrete, a bar chart is best; distances between the bars demonstrate the discontinuous nature of the data.
Histogram: for continuous data where groups can be easily formulated; midpoint of the data is displaying on the x-axis with the frequency as the latitude of the data point.
How would you present the summary of two or more variables?
Box plot: allows for easy comparison of the dispersion and location of each variable.
Multiple line graph: easy to digest and is commonplace is statistical displays.
How would you present the comparison of two variables?
Scatterplot or multiple scatter plot (groups are denoted by different symbols). A line can be added to show variance over a temporal scale e.g. temperature; or a trend line can highlight the relationship between the two variable under investigation.
What are standard experimental designs?
Control, procedural, experimental, temporal and statistical designs all need to be thought through. Basic formats to control for bias or spatial factors that may confound an experiment include a Latin square, blocking and stratified random treatment.
What is the difference between nominal/string data and scale data?
Nominal data has alphabetic characters instead of numerical codes whereas scale data comprises of observations that can assume any numerical value.
What are the two tabbed screens on IBM SPSS 24?
Variable view and data view.
What is a common cause of issues in SPSS?
When scale values are set to nominal or vice-verse.
How would you plot a histogram displaying two variables e.g. the variance of height in females compared to males?
Graph>chart builder>select histogram>select population pyramid>drag height into the variable box>drag gender into the split variable box.
How do you obtain basic descriptive stats?
Analyse>Descriptive statistics>Descriptives>select default settings in options menu plus ‘variance’.
What’s the best way to display the mean and 95% CI of multiple datasets for comparison?
Present as a simple error bar.
Graphs> Chart builder.
Display 95% error bars.
Describe how you would conduct an independent-samples T-test, the null hypothesis and which aspects of the output are important.
Analyse> compare means> Independent samples T-test
Select the variable and drag into the variables box, select the grouping name and drag into the grouping box.
The null hypothesis: That the two samples means come from a population with the same true mean (i.e. the two sets of data are the same).
Group statistics scribe the central tendency of the data and spread; the percentage of deviation away from the mean value can be deciphered.
The output shows the Levene’s test, t-test value, df’s, and Sig(2-tailed); which ultimately deciphers if the null is going to be accepted or rejected.
The 95% confidence interval will generally span 0 if the output is non-significant.
How do you decide between the conduction of a parametric or non-parametric test?
Create a histogram of collected observations to see if the data conforms to the principles of a normal distribution.
Why would you use a Mann-Whitney U-test and not an independent-samples T-test? Describe how you would conduct the test.
You would use the Mann-Whitney test when the data is non-parametric and unpaired; it also performs highly with large sample sizes.
Analyse> nonparametric tests> independent samples.
In the fields tab drag the variable in, into the test box drag the group in. In settings check customise tests and then Mann Whitney U (2 samples).
When would you choose to utilise a paired-samples T-test and how would you conduct it?
Used when an individual has been sampled twice e.g. over a time period or if a cloned individual has been exposed to two different treatments.
Analyse> compare means> paired samples T-test.
Drag both groups into the paired variables box.
Note, in the variable view of SPSS the data must be played out side by side.
What is the non-parametric equivalent to the paired-samples T-test and how would you conduct it?
Wilcoxon Signed Rank Test.
Analyse> nonparametric tests> related samples.
Drag both groups into the test fields box and select the relevant test in the settings test.
How would you conduct a one-sample comparison?
Method 1: One-sample T-test (parametric data).
Analyse> Compare test> One sample T-test.
Drag the variable into the test variable box and set the test value to the relevant value.
Method 2: Wilcoxon Signed Rank Test (non-parametric data).
Analyse> nonparametric> one sample.
Drag the variable into the test fields box and choose the relevant test in settings; alter the hypothesised median to the desired value.
