Statistics - Part 2

Yesterday we spent some time looking at nominal, ordinal, interval and ratio levels of measurement. The key thing here is for you to be able to spot when the data is ordinal or only nominal.

We looked at the Wilcoxon Signed Ranks test - this is used to decide if scores in a repeated measures design experiment are significantly different. The data needs to be at least ordinal, because you have to rank the differences from smallest to largest, then add up the ranks of the positive differences and the ranks of the negative differences. If the smaller of these is less than the critical value (from the table) then the result is significant - in other words only a few results went 'the wrong way' and these only by a bit.

For Monday, make sure you have completed the Using the Sign Test sheet.

Also - gather some experimental data on open and closed passes. An open pass is when a person squeezes past someone else with their front to them - a closed pass is when they put their back to the other person. Create a tally chart for male and female passes and record the number of open and closed ones you observe for each.

Also have a go at the Mann-Witney U Test on the three-page sheet (on which you completed the Sign Test and Wilcoxon). The tricky bit here is that you have to rank the actual scores (NOT the differences between the scores - as this is and independent groups design the scores aren't actually in pairs), and you have to do this AS ONE GROUP meaning that you should get up to 20.

Statistics lesson 1 - the sign test

Here is the presentation from today's lesson.

The key idea to take from this is that when we have quantitative data, we can use a statistical test to see how likely the results were to have happened simply by chance. This is the same thing as the chance of the null hypothesis being true given what happened in our experiment.

Scientists choose a 'level of significance' before starting an experiment. This is the probability of the results happening by chance alone below which they will reject the null hypothesis, and conclude that their alternative hypothesis is true. In psychology 5% (or p<0.05) is usually chosen - this means that if there is a less than 5% chance of the results happening by chance alone then the result is considered to be significant - down to whatever the effect being investigated is.

The 5% level gives a balance between the two types of error - false positives (Type I) and false negatives (Type II). Sometimes the 10% level is used - this means there is a greater chance of a false positive (saying you have found something when you haven't) but you are less likely to have a false negative (saying nothing was going on when really it was). When it's important to avoid a false positive a lower level e.g. 1% or 0.1% is used - but this means that a false negative is more likely.

Here is the 'using the sign test' homework sheet - for Thursday, along with the sign test section on the other sheet I gave out.

Final Unit 2 revision...

Tuesday at 12.20pm in R23!