|
Post by victoria on Sept 4, 2019 19:53:16 GMT
Hi, I am designing an experiment which will have three conditions (a control, test a and test b). I would like to allocate sample across gender and age so that I end up with a broadly matched samples within each condition. Is there any way of using survey questions up front (age, gender) and then using the responses to those to allocate people to a specific 'block' using an algorithm that strives to deliver matched samples? This sounds complicated so I'm prepared to be disappointed but I also know that I have very basic coding skills, so there may be something clever that can be done that I have no idea about!
If this is not something I can achieve within a single piece of coding, I have assumed I will have to run two stages - first, a survey in which I will collect people's age and gender in order to balance the samples and then manually allocate them to each sample group. But I would then need to send out a second link to have them complete the experiment. If I do this, is there a way of linking their original survey answers to their subsequent experiment responses so that I am able to recombine the data into a single set for analysis?
Any help on this very much appreciated!
|
|
|
Post by PsyToolkit on Sept 5, 2019 10:45:58 GMT
Hello Victoria,
Thank you for your question. It is a good question, but there is no straightforward way of doing this.
I can see why you think of doing this, but then, if you would get sufficient numbers of participants, and if the allocation over the three different possible conditions participants might be allocated to is random, you'd expect somewhat similar numbers.
I aggree that your two-stage approach is the easiest to set up currently. That makes a lot of sense, that is how I would suggest my students to run it.
Best,
Gijsbert Stoet
|
|
|
Post by victoria on Sept 9, 2019 19:55:40 GMT
Thank you Dr Stoet. If I am running the survey in two stages, is there a way of combining data from each of the two surveys into a single dataset for the purposes of analysis? I'm not sure how I would go about matching up people's initial classification data with their responses to the main survey if they do them at different times in different links?
Thanks, Victoria
|
|
|
Post by PsyToolkit on Sept 10, 2019 20:39:01 GMT
Yes, that is a good point. There are two different ways of doing this. Option 1: If you work with a small set of participants (say 10 or so), then you could simply have a different survey (and thus different survey link) for each participant. I have done this in the past in which my participants had to do multiple sessions of the same experiment. Each participant thus had a unique survey link, and this made it easy for me to manage. The only thing that is a bit more work is that you need to set up each survey new. That can take 10 minutes per participant (depending on how many questions).You can copy and paste, of course, but still, it is extra work. Option 2: You can send via email each participant a unique ID (or you can even use PsyToolkit to give people a unique id at the end of the session). Then each following session they need to enter that ID. The only drawback of this solution is that people might forget or mistype it.
So in short, I would go for option 1 if possible, but if you have many many participants, option 2 is easier to set up.
Hope that helps.
|
|
|
Post by victoria on Sept 10, 2019 22:07:59 GMT
Excellent, thank you!
|
|
|
Post by Vishnukant on May 24, 2020 5:53:04 GMT
Yes, that is a good point. There are two different ways of doing this. Option 1: If you work with a small set of participants (say 10 or so), then you could simply have a different survey (and thus different survey link) for each participant. I have done this in the past in which my participants had to do multiple sessions of the same experiment. Each participant thus had a unique survey link, and this made it easy for me to manage. The only thing that is a bit more work is that you need to set up each survey new. That can take 10 minutes per participant (depending on how many questions).You can copy and paste, of course, but still, it is extra work. Option 2: You can send via email each participant a unique ID (or you can even use PsyToolkit to give people a unique id at the end of the session). Then each following session they need to enter that ID. The only drawback of this solution is that people might forget or mistype it.
So in short, I would go for option 1 if possible, but if you have many many participants, option 2 is easier to set up.
Hope that helps.
I have a large survey, I want to break it in to sections. After, completing a sections I want to ask participants whether they want to continue or not. if they say yes, the survey should go to next section, otherwise if they say later a code should be generated, and when they do the next session later, the generated code, should automatically become their id. So that its easy to math response from same participant. Is there any algorithm for this. As I see the quoted answer is from sept 2019, is there any update on the issue. Or multiphase, survey is not possible?
|
|
|
Post by Vishnukant Tripathi on May 24, 2020 6:10:17 GMT
Yes, that is a good point. There are two different ways of doing this. Option 1: If you work with a small set of participants (say 10 or so), then you could simply have a different survey (and thus different survey link) for each participant. I have done this in the past in which my participants had to do multiple sessions of the same experiment. Each participant thus had a unique survey link, and this made it easy for me to manage. The only thing that is a bit more work is that you need to set up each survey new. That can take 10 minutes per participant (depending on how many questions).You can copy and paste, of course, but still, it is extra work. Option 2: You can send via email each participant a unique ID (or you can even use PsyToolkit to give people a unique id at the end of the session). Then each following session they need to enter that ID. The only drawback of this solution is that people might forget or mistype it.
So in short, I would go for option 1 if possible, but if you have many many participants, option 2 is easier to set up.
Hope that helps.
I have a large survey, I want to break it in to sections. After, completing a sections I want to ask participants whether they want to continue or not. if they say yes, the survey should go to next section, otherwise if they say later a code should be generated, and when they do the next session later, the generated code, should automatically become their id. So that its easy to math response from same participant. Is there any algorithm for this. As I see the quoted answer is from sept 2019, is there any update on the issue. Or multi-phase, survey is not possible?
|
|
|
Post by PsyToolkit on May 26, 2020 22:21:06 GMT
I think you can do this. Ultimately, you can give people a random code, either using the one that is built-in end code or you can create your own random code.
You would need to manually match the resulting lines in your spread sheet or write your own code for the matching of the data in your spreadsheet. You can do that with a variety of free software such as one of the high level scripting languages (ruby/python, etc) or R.
|
|