I'm designing an experiment with two condition. Cond A with no manipulation of the independent var and ConB with manipulation of the independent var.
The results are measurement taken a certain time intervals (could be up to 500).
As result I have 2 different measurement (Cond A and Cond B) that vary over time. For instance x-axis = time-stamp, y-axis = frequency of observation x
I would like to compare both measurements (A and B) with each other so that I can say something about my H0 and H1. Do both measurement differ enough because of the manipulation?
Looks like you have longitudinal data analysis. You can use repeated measurements. So your hypothesis will read something like this:
"Y for cond B is different from Y for cond A over time". Most software packages accommodate repeated measurements.
Jenny Kotlerman
www.****************************.com
Thanks for the reply. I need to look into that...
Maybe this example wil give you some more information.
Take for instance
http://cloudalicio.us/tagcloud.php?u...ot.org&x=0&y=0
You can see how the tag ratio changes over time. What if I want to show that for a certain tag (for example 'news') my manipulation has an effect on the evolution of that tag over time.
So one group gets a manipulation and the other not. Then I want to compare both curves of the tag 'news' in both conditions. I want to see if my manipulation has a significant effect on the tag evolution.
so you want to show that the weight for tag news after manipulation is different from the weight for tag news without manipulation. I am assuming that you will have two groups (or two tags that are otherwise similar) where one gets the manipulation and the other one doesn't. You measure both tags (A and B) before the start of the study. Your tag that is not getting any manipulation will be used as your control. The idea is, that if both tags change (for example increase) over time, then the effect of time can be separated from the effect of manipulation. If your tags are randomly selected then they should be the same at baseline and they will be different after the end of the study, if the manipulation had an effect.
Jenny Kotlerman
www.****************************.com
Cond A = Control group with no manipulation
Cond B = Manipulation group
Test person are randomly assigned to both conditions. The tags are not randomly selected -> test persons will add tags during the experiment based on their judgments. So we start with 0 tags...
This means that the tags can not measured before testing. People in both conditions tag the same content. Afterwards I want to compare the evolution of the differnt tags over time. What do I want to know : has the manipulation influence on the way people tag?
The problem is ; How do I compare this? For each tag there will be 2 curves : one of cond A (control group) en one of cond B (manipulation group). If all goes well these curves will differ. But is this difference between curves significant so that we can say that it is the result of the manipulation?
if there are two curves for each tag then you compare those two curves with each other. If the difference is significant then you can say that manipulation changed the outcome
Jenny Kotlerman
www.****************************.com
I just compare them? What statistics can I use to that? These are timeseries... I think I can't just compare their means...
no you need to use repeated measures analysis. What statistical software are you going to use?
Jenny Kotlerman
www.****************************.com
Maybe I'm going to use SPSS for it.
Is this also a possibility?
http://en.wikipedia.org/wiki/Time_series
Comparing two curves over time? If this isn't an option...why not?
Posting Permissions
|
|
Reply With Quote