SL Psychology IA - Descriptive statistics for binary data set?

[blank-user]

So for the SL Psychology IA, our teacher has told us that we must use purely descriptive statistics for our data, and that inferential statistics (such as tests for the validity of results) are only for HL students.

The problem is that the research I decided to replicate deals with a binary data set. As in, there are only two outcomes (correct or incorrect). 

I am able to find the mean for a measure of average and that makes sense. However, the problem is that apparently we need to include a measure of spread too... For a binary data set, measures of spread such as variance or standard deviation does not really mean anything.

I guess there are two questions I'm asking:

- Are there any suggestions for what I should do for measure of spread?

- Will I get marked down for including inferential statistics if I feel like it will contribute to my discussion?

[blindpet]

Which research are you replicating? What were your independent and dependent variables?

You can still get the spread by doing the average of the data (it will be between 0 and 1) so you'll get a standard deviation, of course showing a table of the distributed yes and no will be much easier for the user to read, that table shows the spread of data.

With binary choice data you can use chi squared or logistic regression but they are beyond the scope of psychology SL I believe.

[blank-user]

13 minutes ago, blindpet said:

Which research are you replicating? What were your independent and dependent variables?

You can still get the spread by doing the average of the data (it will be between 0 and 1) so you'll get a standard deviation, of course showing a table of the distributed yes and no will be much easier for the user to read, that table shows the spread of data.

With binary choice data you can use chi squared or logistic regression but they are beyond the scope of psychology SL I believe.

Wason's selection task, with the independent variable being the change in context of the logical problem, and the dependent variable being the number of participants answering the problem correctly.

I understand that I am still able to calculate the standard deviation. However, it is just that it has very little meaning as there isn't really a spread of data. Measures of spread work best when you have a range of values, but in my case, there are only two possible values (0 and 1).

Anyway, that's one matter.

 

The other is whether I am able to use statistical tests in my IA or not... I was planning on using Fisher's test to look at the significance of the difference in result between the two contexts. 

[blindpet]

Fisher's exact test would be good and is what is used elsewhere for the same research http://www.fitelson.org/probability/wason.pdf

You shouldn't lose marks but it will take up words so you should be mindful of that as you will not get marked up for including Fisher's test (it's not part of any assessment criterion). The table will show the distribution of scores which is the best you will get for your nominal data.