Self-reporting Test Feedback Thread

Thread posts

See all posts in this thread below.

  • User
  • Message
  • Actions
    • Published: 12-31-2014 05:03 am
      Updated: 01-02-2015 04:31 pm
    • This thread will be used for feedback related to the new self-reporting test released today, we are doing the final touches now.

      I think feedback related to the test can be divided into different categories:

      1. Does the types exists. This should be motivated by empirical sources.
      2. Does the type traits relate to the types. The traits should be directly derived from the types and also motivated by empirical sources.
      3. Does the self-reporting questions measure the traits. Questions should be possible to motivate by the traits.
      4. Is the calculation of the self-reporting test done correctly. The mechanics of test should be scientifically valid.
      5. Is any question too ambigious.
      6. Is any feature missing from the test.

      We are working on creating temperament descriptions in the test results. We will also analyze the feedback and see what can be improved, we have a few items we know can be improved already.
    • Published: 01-09-2015 03:49 am
      Updated: 06-25-2015 02:52 am
    • 4. Is the calculation of the self-reporting test done correctly. The mechanics of test should be scientifically valid.

      Premises

      Scores per question is like this:
      -30 = Strongly disagree
      -15 = Disagree
      0 = I didn't understand the question
      15 = Agree
      30 = Strongly Agree

      24 distinct traits.

      All questions relate to exactly one trait.

      Four questions per trait.

      All distinct 8 types have a distinct set of 12 traits.

      All results consists of 96 answers.

      Method

      The idea is that people report correctly to 75%, so out of four questions per trait the person reports correctly 3 of them. So the most divergent answer (absolute distance from mean) per trait is replaced with the mean answer for the other answers. So answering 30, 30, 30, -30 should instead result in the answers 30, 30, 30, 30. Another example is 30, 15, 30, -15 which should result in 30, 15, 30, 25

      Mean answer per trait is calculated.

      Answers for "I didn't understand the question" is replaced by the mean answer for the specific trait.

      Total score per trait is calculated.

      Total score per type is calculated by summing up the total score of the traits that belong to each type.

      Variance and standard deviation per trait is calculated.

      Norm of standard deviation per trait is calculated.

      Norm of standard deviation per type is calculated by summing up the absolute values of the standard deviations for all traits per type.

      Balanced score per trait is calculated as described above

      Balanced score per type is calculated = total score per type - standard deviation norm per type.

      Presentation

      Types are sorted by balanced score in descending order.

      Type result is the type with highest balanced score.

      Traits are sorted by balanced score in descending order.
    • ErikThor likes this post.
    • Published: 01-20-2015 05:14 am
      Updated: 01-24-2015 05:01 am
    • 5. Is any question too ambigious.

      This seems to be the biggest issue with the current test. People have trouble knowing what context the questions is asked in and many people request concrete examples, like what kind of things are meant by tools?

      I think we need to add both examples of each question and also context for the words used. So we need to expand questions. Also we need to replace some questions which are hard to understand in general.

      Questions which asks about most should be changed to exclude the word 'most' or supply a alternate scenario to compare with.
    • Published: 03-26-2015 02:44 am
      Updated: 03-26-2015 06:54 am
    • I think the self-reporting test is fairly accurate for most people. One of the most interesting finding is how hard it is to quantify traits which are complex in nature, it's hard to define the questions in a way that people get what you mean.

      Based on study of correlations from the traits of the test it seems the Cognitive-Affective scale is mostly related to Execution-Processing dichotomy. Some say they relates it to the Mechanic-Organic dichotomy but most don't.

      I think we should make some adjustments to the type models, I'm thinking something like this instead, adding anti-correlations as well:

      Traits per type:

      AME
      Correlated: AE, AG, GE, FP, ME, MP, OP, CP
      Anti-correlated: CE, CG, FE, GP, OE

      AOE
      Correlated: AE, AG, GE, FP, OE, OP, MP, CP
      Anti-correlated: CE, CG, FE, GP, ME

      CME
      Correlated: CE, GE, CG, FP, ME, MP, OP, AP
      Anti-correlated: AE, FE, AG, GP, OE

      COE
      Correlated: CE, GE, CG, FP, OE, OP, MP, AP
      Anti-correlated: AE, FE, AG, GP, ME

      AMP
      Correlated: AP, GP, AG, FE, OE, OP, MP, CE
      Anti-correlated: CP, FP, CG, GE, ME

      AOP
      Correlated: AP, GP, AG, FE, ME, MP, OP, CE
      Anti-correlated: CP, FP, CG, GE, OE

      CMP
      Correlated: CP, GP, CG, FE, OE, OP, MP, AE
      Anti-correlated: AP, FP, AG, GE, ME

      COP
      Correlated: CP, GP, CG, FE, ME, MP, OP, AE
      Anti-correlated: AP, FP, AG, GE, OE
    • ErikThor likes this post.
    • Published: 03-26-2015 08:06 am
    • Ok I changed the traits to type association and it seems to give better results. Most results seems to get same type but different balanced score.

      I think a lot of results which looks like this really give evidence for that the self-reporting test actually finds real differences in personality.
    • Published: 04-03-2015 12:47 pm
    • I have found some guidlines for the self-reporting test.
      * Use questions like today (avoid fixed choice like MBTI), this to be able to provide a correlation matrix between traits, otherwise the statistics is tainted because people are forced to pick between two traits were it can't be excluded people prefer both to a equal degree
      * Analyze answers per question and try to reach a state were each question gets a normal distribution in response
      * Summarize total score per trait and analyze this to reach a state were each trait gets a normal distribution in scores
    • anders likes this post.
    • Published: 04-06-2015 06:33 am
      Updated: 04-07-2015 09:33 am
    • By changing the traits per type a bit we can reduce the number of questions by 16 soon. By removing OP and MP since it seems all people have these traits we get this list instead:

      New traits per type:

      AME
      Correlated: AE, AG, GE, FP, ME, CP, CF
      Anti-correlated: CE, CG, FE, GP, OE, AF

      AOE
      Correlated: AE, AG, GE, FP, OE, CP, CF
      Anti-correlated: CE, CG, FE, GP, ME, AF

      CME
      Correlated: CE, GE, CG, FP, ME, AP, AF
      Anti-correlated: AE, FE, AG, GP, OE, CF

      COE
      Correlated: CE, GE, CG, FP, OE, AP, AF
      Anti-correlated: AE, FE, AG, GP, ME, CF

      AMP
      Correlated: AP, GP, AG, FE, OE, CE, CF
      Anti-correlated: CP, FP, CG, GE, ME, AF

      AOP
      Correlated: AP, GP, AG, FE, ME, CE, CF
      Anti-correlated: CP, FP, CG, GE, OE, AF

      CMP
      Correlated: CP, GP, CG, FE, OE, AE, AF
      Anti-correlated: AP, FP, AG, GE, ME, CF

      COP
      Correlated: CP, GP, CG, FE, ME, AE, AF
      Anti-correlated: AP, FP, AG, GE, OE, CF
    • Published: 04-07-2015 01:29 pm
    • Ok it's now possible to select specific traits to test, this will improve testing and development a lot. I will add buttons to select all required for a type result, all traits and specific traits.
    • Published: 04-22-2015 03:44 am
    • I don't think everyone has OP and MP. I think the results were off because we formulated the questions poorly. At least, we can't know that for sure yet.
    • Christian likes this post.
    • Published: 04-22-2015 03:48 am
    • Yes, that's possible. Maybe our definition was off.
  • User
  • Message
  • Actions