How to do your own experiments in psychology and education
7.4 Organize and code your data
Congratulations! Now you have a big stack of papers or measurements – real data!
Make backup copies. If you collected measurements with a computer or other instruments, then the very, very first thing to do is back everything up:
This is Salkind’s (2006) fourth Commandment of data collection.
Get organized. Now make sure that:
Separate out all of the consent forms and put them in a manila envelope to archive. You won't neet them for the rest of the experiment.
For each experimental condition, use a separate manila envelope and put all of the data for that condition in the same envelope. Write on the envelope, on each response sheet, and each background questionnaire which condition it is from.
BTW, yes, you can write on your participants’ response sheets. It’s your data, as long as you don’t change their responses. You may want to write group numbers, analysis codes, or preliminary counts on each sheet.
Coding your data
Your next step is to code the data: the information from the background questionnaire, the factors or independent variables, and the responses – your dependent measures – to make it easier to manipulate in the statistical analysis programs.
Data coding for dependent measures. The next step is to score or code the data. What score will this participant get for each of the things that you want to measure?
Consistency is why you really need written instructions to follow and not just your intuition: if you follow explicit instructions then you will be more consistent. With written instructions, you can get help to do the coding and it’s likely (not guaranteed!) that the different coders will almost do the same thing as you would.
In the easiest case, you will have a list of acceptable or “correct” responses and you will just count up the acceptable responses for each participant. Say you have a test questionnaire with 20 multiple-choice reading comprehension questions. For each question, you can simply check the participant’s response again your list of “correct” responses. This is an easy method, but it doesn’t provide a lot of information.
Data coding for background questionnaires and factors. You have to do something similar with the information from the background questionnaires. At least, you should do it for the parts of the questionnaire that you might want to test during analysis. You have to code the parts of the questionnaire that measure your factors; any other items are optional -- you can always go back and check them later.
First you will need to devise a key so that we all know what your numbers mean, as in the examples below. We’re treating each question on the questionnaire as if it were a factor and the different answers can be seen as levels of that factor. The coding, at least, is done in exactly the same way.
Gender (1 = Feminine; 2 = Masculine)
Major (1 = Psychology; 2 = Other)
First Language (1 = English; 2 = Other)
In this way, you enter the codes into the spreadsheet.
7 2 1 1
This example means that this participant (#7) was a male (2) psychology major (1) who’s first language is English (1). Then, in the next column, you'll write how many items he recalled or whatever you happen to be measuring. The next row will have parallel information for the next participant, and so on until you fill out a spreadsheet like the one here.
Recall tasks. If you ask participants to read something and then write down everything that they can remember, the task is called free recall and their responses are called free recall protocols. The participants basically write anything that they want. Similarly, if you ask the participants to answer in writing short-answer or essay questions about what they read, then it's called a probed recall task (because you probed their memory with questions).
This kind of data is more complex to analyze, but itcan provide you with very interesting information about how people think while they're reading. There's more on how to code this type of data here.
Coding your factors. Finally, you will do the same thing for your factors / independent variables: give each level a code and enter the codes for each participant in the spreadsheet.
For example, the participants who heard music might get a 1 in the music column and the participants who did the experiment in silence might get a 0 in the music column.
Make backup copies. If you collected measurements with a computer or other instruments, then the very, very first thing to do is back everything up:
| 1. Make an extra, backup copy of all of the data files, | |
| 2. Double-check that everything was copied correctly, | |
| 3. Make sure that the backup is write protected (can’t be erased), and | |
| 4. Hide the backup copies someplace safe. |
This is Salkind’s (2006) fourth Commandment of data collection.
Get organized. Now make sure that:
| a) all of the data for each participant is together in the same place, | |
| b) you’re absolutely sure that each piece is from that participant, and | |
| c) each piece has that participant’s code on it. |
| Reality check: if someone (crazed supervisor, mischievous child, enraged significant other, …) threw all of your data out the window, could you put it all back together again? If not, then your data isn’t organized yet. |
Separate out all of the consent forms and put them in a manila envelope to archive. You won't neet them for the rest of the experiment.
For each experimental condition, use a separate manila envelope and put all of the data for that condition in the same envelope. Write on the envelope, on each response sheet, and each background questionnaire which condition it is from.
BTW, yes, you can write on your participants’ response sheets. It’s your data, as long as you don’t change their responses. You may want to write group numbers, analysis codes, or preliminary counts on each sheet.
Coding your data
Your next step is to code the data: the information from the background questionnaire, the factors or independent variables, and the responses – your dependent measures – to make it easier to manipulate in the statistical analysis programs.
Data coding for dependent measures. The next step is to score or code the data. What score will this participant get for each of the things that you want to measure?
| Be sure to score or code the data in the exact same way for each and every participant in all of your conditions. Don’t try to make up a coding procedure as you go. |
Consistency is why you really need written instructions to follow and not just your intuition: if you follow explicit instructions then you will be more consistent. With written instructions, you can get help to do the coding and it’s likely (not guaranteed!) that the different coders will almost do the same thing as you would.
In the easiest case, you will have a list of acceptable or “correct” responses and you will just count up the acceptable responses for each participant. Say you have a test questionnaire with 20 multiple-choice reading comprehension questions. For each question, you can simply check the participant’s response again your list of “correct” responses. This is an easy method, but it doesn’t provide a lot of information.
Data coding for background questionnaires and factors. You have to do something similar with the information from the background questionnaires. At least, you should do it for the parts of the questionnaire that you might want to test during analysis. You have to code the parts of the questionnaire that measure your factors; any other items are optional -- you can always go back and check them later.
First you will need to devise a key so that we all know what your numbers mean, as in the examples below. We’re treating each question on the questionnaire as if it were a factor and the different answers can be seen as levels of that factor. The coding, at least, is done in exactly the same way.
Gender (1 = Feminine; 2 = Masculine)
Major (1 = Psychology; 2 = Other)
First Language (1 = English; 2 = Other)
In this way, you enter the codes into the spreadsheet.
7 2 1 1
This example means that this participant (#7) was a male (2) psychology major (1) who’s first language is English (1). Then, in the next column, you'll write how many items he recalled or whatever you happen to be measuring. The next row will have parallel information for the next participant, and so on until you fill out a spreadsheet like the one here.
Recall tasks. If you ask participants to read something and then write down everything that they can remember, the task is called free recall and their responses are called free recall protocols. The participants basically write anything that they want. Similarly, if you ask the participants to answer in writing short-answer or essay questions about what they read, then it's called a probed recall task (because you probed their memory with questions).
This kind of data is more complex to analyze, but itcan provide you with very interesting information about how people think while they're reading. There's more on how to code this type of data here.
Coding your factors. Finally, you will do the same thing for your factors / independent variables: give each level a code and enter the codes for each participant in the spreadsheet.
For example, the participants who heard music might get a 1 in the music column and the participants who did the experiment in silence might get a 0 in the music column.
| Triple check everything that you put into your spreadsheet. Errors of data entry make the results very difficult or impossible to interpret. Another advantage of working with a partner is that you have someone to help your review data entry. |
| Read this topic next to review data collection: | 7. Collect your data |
About
© 2009 Created by Mike Dillinger, PhD on Ning. Create a Ning Network!
