Search results

  1. hlsmith

    Trend analysis

    What study, did you actually intervene some how or what would have driven clinicians to order MRIs faster? You may want to rerun the model without the outlier just to make sure the outlier isn't pulling the line up, but likely you won't see a difference.
  2. hlsmith

    Should i use McNemar's for a matched-pair study?

    What is the standard deviation of these values, the issue would be that say + 2(std) may be something like 30, which would not be possible if the top value has to be contained within 100. Are the scores a percentage, if so you may want to investigate beta regression.
  3. hlsmith

    Trend analysis

    Yes, but I imagine you had more than 30ish MRIs performed over 14 year period? Are these the raw delay values for each MRI conducted? Also, you would want to confirm the high plot was really an outlier and not a calculation or data recording error. Next you would look to plot a line to these...
  4. hlsmith

    To assess meaningfulness between an independent variable and a dependent one, is just the p value alone adequate and/or the correlation?

    Are you using linear regression? If so, look at magnitude and direction of slope along with 95% confidence interval.
  5. hlsmith

    Jupyter notebook

    When I installed a jupyter notebook on my computer via anaconda, it auto-created a bunch of folders in it. Consisting of Music, photos, desktop, etc. Not sure what's up with that. Also, it tells me I can't delete some of the folders since they aren't empty, but they appear to be empty. Anybody...
  6. hlsmith

    Should i use McNemar's for a matched-pair study?

    What are the average scores? If the aren't around 50 or have quite a bit of dispersion, linear regression approaches can have problems. This is due to value being contained within 0 and 100.
  7. hlsmith

    Should i use McNemar's for a matched-pair study?

    Does each patient have a single match?
  8. hlsmith

    How can I control the format of the PCA analysis output in R?

    Why would it look the same for two different datasets? That doesn't make sense.
  9. hlsmith

    Bias correction

    You get prediction and have observed data, so you have residual or error data. Unless there is a pattern in the errors I don't know what you correct for, you just have bad fit. If you have a pattern you can try to transform data or tweak the model!
  10. hlsmith

    Linear Probability Model

    Referring to the logistic regression assumption of linearity in the logit, which isn't a concern when predictor is categorical.
  11. hlsmith

    Proportionate vs absolute count for analysis

    Your description is very vague. What procedure are you trying to use? Are you trying to use a Chi-sq test? Also, what program are you using? Also, I can't follow what you are looking at, how does 20 species translate to 15.5%; also if you don't use count based values, how does the program know...
  12. hlsmith

    Linear Probability Model

    I didn't exactly follow your quote, but is their lack of concern with a binary variable being you the slope between the two values is just a straight line. The same lack of concern in regards to the linearity of the logit for binary predictor versus a continuous predictor.
  13. hlsmith

    Alpha 0.1 and power 75% is it acceptable

    jm, You could provide more information. Given your posts, I imagine you conducted your study and failed to reject the null. Then conducted a sample size calculation which appeared provide weak alpha/beta. Though those values can be traded off in decrease one and increase the other.
  14. hlsmith

    Bias correction

    Agree with Maarten, please describe your issue so we can make informed recommendations. Bias correction can drastically vary from external to internal data adjustments.
  15. hlsmith

    Time Series Analysis

    Great questions. Are you expecting us to answer them for you?
  16. hlsmith

    Hierarchical linear modeling - minimum number of observations per group?

    I feel there is a formal test to examine if controlling for groupings explains a greater amount of variability, if it doesn't you can use fixed effects with robust standard errors.
  17. hlsmith

    Hierarchical linear modeling - minimum number of observations per group?

    I am sure there are better people out there to answer this question, more knowledgeable, but yeah that seems like too few. To put it in perspective how much more variability can be explained by controlling for the second level. Do you think they have random intercepts and slopes? So with the...
  18. hlsmith

    Chi square effect size

    Why. I could flip a coin a kabillion times and get 0.0001% heads, then flip another coin and get 0.0003% heads. Well I can conclude that the rates are different and hey they might be as the rate converges to the truth with every trial. Is that a big effect. Nah. But the study is well powered...
  19. hlsmith

    Software for Simple Linear Regression

    Pretty much. I believe it may have a model fit option for quadratics - I can't recall if that included smoothing. Though I think data can just be transformed to incorporate smoothing. R and Python are software options as well, but traditionally thought to be more technical.
  20. hlsmith

    Minimise chance of making type 1 error, without reducing alpha?

    Minimize measurement errors, minimize selection bias, control for positive confounding, control for mediating effects - that is given these things exist and functionally serve to bias effects away from the null.