Single Case Studies - ARMS (neuropsychology)

If you have norm-data available, you assess a patients data using norm-tables such as this one:

You correct for age and education, and correct the score for this, using the table. Then you look in what scale the patient’s score is:

And you look what percentile the scale lies in:

But what if your patient has motor impairment or couldn’t draw or speak well? Then you can’t use this test. Options:

• Use a different test
• Make a new test. But: there are no norm data. So make your own!
  
  • Good match: same age, gender education level
  • Good control group: no neuropsychological disorders
  • Sample size?

Imagine you have a mean and SD from a population and you want to know if your patient’s score differs significantly from this? Use a z-test:

Fill this in:

If your Z score is lower than the critical z score, it is significant! If you do a one-tailed test: it is -1.65.

From the z-score you can also calculate the p-value (in excel):

• Z=normsinv(p)
• p=normsdist(Z)

If your own control group is the red line, but the population/normscores is the black line, you will get different results. Your patient (green) does not differ from the actual population/normscores, but does from your own controlgroup. This is a type 1 error/false alarm.

How to score a ROCF test:

1. Score the test using test manual, and assign points.
2. Correct the scores
3. Translate to scale
4. Translate to percentile

A different statistical test (than z-test) you can use to know the difference between your control group and the population is the one-sample t-test.

• Z-test: we treat the stdev and mean of the sample as population representative. This results in a bigger type 1 error.
• One sample t-test: uses sample stdev and takes into account number of participants. Lower chance of type 1 error.

With the modified t-test you can compare one participant to the control group:

Test significant --> person deviates from control group.

In a monte carlo simulation, you test the chance of a type 1 error. The z-test has more type 1 errors in low N's than the modified t-test, also if it is skewed.

When conducting a t-test, look at the critical t-value table to see if your result is significant. If you work with a participant, you look at the one tailed significance level. The df (degrees of freedom) is the amount of control participants – 1. The critical t-value that you read in the table tells you: all the absolute values higher than the critical t-value are significant.

In the program SINGLIMS_ES you can also compute the t-value.

Dissocation: a patient deviates in one task, but does not in another task. Then there could be a dissociation between the two tasks.

The criteria for dissociations are:

1. Patient differs from controls on task X
2. Patient doesn’t differ from controls on task Y
3. Patient’s performance on task X and Y differ

Before you compare two tasks (to check criteria 3), you have to standardize the scores. The revised standardized difference test lets you compare two tasks with standardized scores:

You can compute this formula in the Dissocs_ES program. 

Report your results the right way!

Questions? Let me know in the contribution section!

Follow me for more summaries on statistics!