I have recently as part of my bachelor project at my university conducted a two-alternative forced choice test. I have looked at ways to say something with significance about which images (the test participants were given a set of 4 pairs of images, for each pair they had to choose one). While testing is emphasized in my education, I do not have enough knowledge about analysis of the results I have now.

How do I go about figuring something significant out about the independent variables that I am manipulating (in separate scenarios)?

Is logistic regression something I could use or are there methods that are easier to understand for someone who has only had one course of statistics?

The data (and edits about the data):

The data consists of rows of test participants and four columns named as image pairs that the participants were choosing between. The values are the test participants' choice (between the two images in the pair)

I have four independent variables, three of them scales and one nominal that I am changing in four scenarios respectively. Each scenario has paths that are chosen "at random" (depending on the last number on the test participant's digital watch e.g. if the time is 13:37, the test participant takes path 7). The reason for the paths is a detail of the implementation in Google Forms to allow all image pair combinations to be presented but not all to every test participant. I would like to find a model for achieving either high probability of one outcome of the dependent binary variable.

  • IV: Amount, Spread, Diff objects, Placement model (nominal)
  • DV: Image chosen or not.

e.i. if I want a high probability of YES/1/ACCEPT/etc answers (to the binary question: "Does the virtual environment seem inhabited to a greater extend than the other image in this pair of images?"). to an image, what need the ratios between the independent variables be?

Here is a link a copy of the spreadsheet containing the data. https://docs.google.com/spreadsheet/ccc?key=0Am7BMwlOEiUNdGJhNU1uQlg5QUtxYW1ZNHhfUHlqS1E&usp=sharing

  • 1
    $\begingroup$ You don't conduct statistical inference by collecting data and then saying "I wonder what might be statistically significant?". You start with questions you want to answer, collect data that well let you answer those particular questions, and then see what it says about those questions. If you collect a bunch of data and then start looking for something significant ... what you're doing sounds awfully like significance-hunting. Instead: What are some interesting questions you want to ask? ... or even more basically "Why would anyone conduct such an experiment?" $\endgroup$ – Glen_b -Reinstate Monica May 21 '13 at 8:19
  • $\begingroup$ Could you describe in more detail what data (what variables of what kinds) you have collected? (Also, since it's for your university study, could you please add the self-study tag?) $\endgroup$ – Glen_b -Reinstate Monica May 21 '13 at 8:22
  • 1
    $\begingroup$ I am not sure we should treat this as a self-study question. A student carrying out his or her final bachelor's project is pretty much in the same situation as a researcher facing a reviewer, writing a grant application or completing a dissertation. It's obviously not a well-defined question straight out of a textbook. $\endgroup$ – Gala May 21 '13 at 9:33
  • 1
    $\begingroup$ @MHA What you are trying to achieve is still unclear. What are those independent variables (what type/how many)? What question did you ask about those pictures (e.g. were participants supposed to rank them or express preferences)? Did everyone see the same pairs of pictures? Was each picture shown only once or several times? And finally and most importantly: What do you want to learn about those pictures or those scenarios? $\endgroup$ – Gala May 21 '13 at 9:36
  • 1
    $\begingroup$ How do the independent variables enter into the design? You still did not answer that question and they do not appear in your data sample. Did you create the pictures to reflect various levels of “amount”, “spread”, “diff objects” and a “placement model” (whatever that is)? Did you somehow measure these variables for each of your pictures? $\endgroup$ – Gala May 21 '13 at 17:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.