Recent content by katxt

  1. K

    Is this is an obvious or tricky question?

    Work out the height of each normal graph at x = 0.7 The likelihood of each distribution is proportional to the heights of the graphs. =NORMDIST(0.7,0,1,0) and =NORMDIST(0.7,1,1,0) in Excel
  2. K

    chi square distribution, degrees of freedom, etc.

    You're quite right, Dason. I should have said the Chi square test, not the chi square distribution.
  3. K

    chi square distribution, degrees of freedom, etc.

    The important point is that the chi squared distribution applies only to count data. It can't be applied to heights, for instance. You count some things and then think "How unlikely is that?" For instance you catch 100 crabs and 66 are female. You say "I expected 50:50. How unlikely is it to...
  4. K

    Calculate the probability that a shipment is good and is rejected.

    Are you sure? Is "the probability that a shipment is good and is rejected" the same as "the probability that it will be rejected given that it is good"?
  5. K

    Test wanted for comparing algorithms

    Q1. That's right. Size is between subjects. Another way of looking at it is a nested anova with instance nested in size. Q2. This needs thinking about. One problem I can see is that the responses are unlikely to be normal because of so many 100% quality responses.
  6. K

    Test wanted for comparing algorithms

    At the simplest level, variables can be either numerical (on a scale of some sort), or categorical. So, algorithm type is categorical. Size can be thought of as a category, (10/small, 20/medium, or 30/large) or as a number on a scale of perhaps 1 to 50, say. It sounds like the former would be...
  7. K

    Test wanted for comparing algorithms

    The structure of your test looks like a two way repeated measures anova. One factor is algorithm and the other is size. If the data is suitable, then looking at the anova interaction will also let you answer questions of the type in part 2 above. Try this and see if the residuals are normal...
  8. K

    Test wanted for comparing algorithms

    It sounds as if the data is matched so a paired test in case 1 (t or Wicoxon) and a repeated measures test in case 2 (repeated measures anova or Friedman) would be siutable. In case 2 you can then use t or Wicoxon as post hoc tests with a suitably reduced critical p value (Bonferroni is common...
  9. K

    Which statistical tests to use?

    Perhaps the next thing you can try is to plot incidents vs passes and see if there is any sort of pattern. Post a graph if you can.
  10. K

    Pathology slides generated between tissue types

    If the rates for the two groups were actually the same, you would have expected to see 1119 in each group. So your question is how unlikely is a split of 1559 vs 679 if the groups are in fact equivalent? This is a chi square test whose p value is best expressed simply as p<0.001. kat
  11. K

    Which statistical tests to use?

    Is there any particular reason to think that the pattern in 2014/15 is different from 2015/16? Life is much easier if you can just lump both years into one list. kat
  12. K

    Non-parametric alternative to Two-way repeated measures ANOVA

    Have you tried transforming the concentration, say by logging it? Or perhaps try a randomization test.
  13. K

    Question about chi-square test and single columns

    I assume that the data is suitable for a chi square test (count data, independent etc). One problem is that of multiple p values when doing several tests which increases the chance of a false positive. You can do a chi square on the 4x3 group as a whole and then, only if that is significant, do...
  14. K

    Am I using the binom.dist properly?

    This doesn't sound like a binomial problem. For a binomial problem you need to be asking things like the probability of getting 5 or more out of some number but there isn't really any "some number" to be out of. How about the Poisson?
  15. K

    What is this formula? Used to convert quarterly basis data to monthly basis

    It looks like some sort of interpolation formula, but I don't see why the third month projection isn't the same as the next quarter projection.