P335 Spring 2001

Instructions for Data Analysis for EEG Projects.

You will have received a text file, which you should open in Microsoft Excel. To do this, you may have to set the file type to .txt. Read it in as a tab-delimited file. If it opens correctly you should see something like this:

In this document, the first column represents the time in milliseconds (thousands of a second). We start recording 100 ms before the stimulus comes on, and so these numbers start at -99 and count up to zero and then to 1100 ms.

The next 4 columns in this example represent the voltage values for the 4 conditions in this experiment. You might have more or fewer. All of the data visible above come from Channel 1. Other channels are represented in groups off to the right.

To graph this data, click and hold on the msec label, and then drag all the way down to the end of the data:

like this:

We record for a total of 1200 ms, which is why this goes up to 1100 ms.

Once you've selected the data, click on the chart wizard button:

which will bring up this dialog box:

Select XY (Scatter) and the lower-right sub-type box.

Click the 'Next' button to get to the next part. Select the Series tab to see something like the box above. By clicking on the Series items in the lower left, you can change the name like I have above.

Once you've renamed all the series, click the Next button:

Now you can add text to the labels. Use something like the labels I've used above.

Click Finish to put the graph in your spreadsheet.

You might want to copy this figure and paste it in several places. That way you can hide some curves to highlight certain comparisons (like just faces vs. inverted faces). Hide a curve by double-clicking on a curve to bring up this dialog box and select 'none' for the line:

Repeat this process for other channels that you are interested in displaying. Only display channels that contain data that demonstrates differences that are interesting or bear on your research question.

You might also double-click the x-axis of your graph and change the scale to go from, say, 0 to 400 ms to highlight the N170 range.

Once you have your graphs, start looking at the conditions, especially the N170, and describe how your different conditions affect the N170's onset and magnitude. Does inversion affect it (if you manipulated inversion)? What does that tell us about what the N170 represents? Come up with a pretty complete statement about how your data address the N170 (or other aspects of your data) and what this means for how faces are processed. Are they processed configurally, and if so, what is the role of the N170?

These are just sample questions you might address; feel free to come up with your own. Use the poster I handed out in class as a guide to the format and types of questions researchers ask.

You can also do other forms of data analysis like a correlation coefficient. Do to this, choose two conditions that you'd like to compare. Then select insert function:

and select correlation:

This will let you select two columns (conditions) to compare:

Click on the red arrows and select first one and then the other column. The correlation coefficient is a measure of how closely two conditions track each other. If you think that the relevant data for the correlation exists only in a certain time range, you could select only the data that come, for example, from between 100 and 350 ms. Use your judgment about which range to include.

The correlation coefficient is a measure of how closely two conditions track each other. For example, if one curve is going up and the other curve is going down in a particular range, they would not be correlated very highly. However, if they are almost on top of one another then the correlation would be higher.

Do the correlations between different conditions within a channel and compare them (sort of like what you did for the face recognition homework assignment). Do you get higher correlations between certain conditions? How do these address your research question?