DSO 530: Logistic Regression in R

Поділитися
Вставка
  • Опубліковано 19 січ 2025

КОМЕНТАРІ • 30

  • @steventhehistorian
    @steventhehistorian 9 років тому

    Excellent explanation, this is EXACTLY what I needed to help me understand this operation in R. Thanks for taking the time to make this!

  • @masudcseku
    @masudcseku 10 років тому

    Thanks, hundreds of web search finally became helpful.

  • @shudharsanmuthuraj1076
    @shudharsanmuthuraj1076 9 років тому

    very very good explanation..appreciate your effort!! keep continuing the good work!

  • @chandrabhatt
    @chandrabhatt 10 років тому

    Thanks for good walk thru to logistics regression.
    Your concluding remarks were awesome "flipping coins might give you better results" :)

  • @ecotrix132
    @ecotrix132 9 років тому +3

    Very well explained! :D I request you to teach different types of regression - Lasso, Ridge, Quantile, Bayesian, Jackknife, etc.

  • @michaelmartin74
    @michaelmartin74 8 років тому

    Is that a Utah State shirt!?? I'm a Ute but always cheer for the Aggies! Your videos are extremely helpful! Thank you for taking the time to make these videos!

  • @lizaleroux5470
    @lizaleroux5470 9 років тому

    Thanks for sharing, this has been MOST helpful

  • @sheidama
    @sheidama 9 років тому +1

    Thanks for the video, but it seems like this doesn't work when the sizes of training and testing data are not equal. It gives a warning message for predict function and then gives probabilities in the size of training data. Could you explain why this happens?

  • @HarmeekHans
    @HarmeekHans 11 років тому +1

    Hi Abbas this is a great video. Are you planning to make a video on goodness of fit for logistic model? I understand you are using misclassification rates using .5 as cut off to get predicted values but that might not be applicable in many areas.

  • @ouhmadlamia9581
    @ouhmadlamia9581 9 років тому

    Excellent explanation, but I have a question concerning the last conclusion. According to the high misclassification rate , you said that the variables are not good. what we should do in this case?
    Many thanks.

  • @JeffJohnsonMrJeffToYou
    @JeffJohnsonMrJeffToYou 10 років тому +2

    This is a great walkthrough! Thanks so much for posting this. I have a question on the creation of test vs training data. Wouldn't it be less biased to randomly select rows? For example use like runif and then sort them? Just curious as I'm studying this now

  • @saidamoaddine1641
    @saidamoaddine1641 9 років тому

    How does the (for) loop looks like if i want to do the logistic regression procedure on 100 different trainingset (so provides me 100 different models) and make predictions based on this 100 models upon one and the same test set?

  • @sureshy5208
    @sureshy5208 10 років тому

    Could you share about the lift chart in R for logistic regression ? I come from a SAS background.

  • @v1kkip
    @v1kkip 8 років тому

    can you create a training video on conditional logistic regression (matched analysis) when there are 1:1 matched case controls, but 100,000 of observations?

  • @My-NaMeS_jEfF
    @My-NaMeS_jEfF 3 роки тому

    Not able to subset Direction by 'test' data set - invalid subscript of type 'list'
    I will subset as Direction[999:1250]

  • @sagarkahane
    @sagarkahane 8 років тому

    superb keep posting videos on r bro please

  • @NB-fb3lr
    @NB-fb3lr 10 років тому

    Can you upload the datasets for your videos? I found one for the first video but the ones on that link don't match the later videos in particular this one, regression, and decision tree analysis

    • @cybergnosis369
      @cybergnosis369 8 років тому

      They can be found in the ISLR package, see here: cran.r-project.org/web/packages/ISLR/index.html

  • @mounisaisiddarthamiddela2199
    @mounisaisiddarthamiddela2199 7 років тому

    Sir, I have one problem with this lecture. When you are converting the data obtained into 0 and 1 after finding out the probabilities,u are considering that when the probability is greater than 0.5 then you are saying that the stock will go up (which is not correct - I think so). Can u please explain?

  • @PramodKumar-cf3tr
    @PramodKumar-cf3tr 8 років тому +1

    I like your tutorial. Its one of best in logistic regression examples.
    1. When i try to use all independent variables eg: I got following warning. Do you know why?
    stock_model = glm(Direction~., data=training_data, family = "binomial")
    Warning messages:
    1: glm.fit: algorithm did not converge
    2: glm.fit: fitted probabilities numerically 0 or 1 occurred
    2. Also, how to identify important variable in logistics regression?
    Thx in advance.

    • @mounisaisiddarthamiddela2199
      @mounisaisiddarthamiddela2199 7 років тому +1

      It may be because, In the dataset, one particular variable for all observations may be same or zero.

  • @dlb8685
    @dlb8685 11 років тому +2

    a) This was very helpful. Thanks!
    b) LOL at the actual model you built. This is not helping me meet my investment goals.

    • @AbbassAlSharif
      @AbbassAlSharif  11 років тому +2

      Hahahha, thanks Dan. flipping a coin might give you better results. LOL

  • @RahulYadavProfile
    @RahulYadavProfile 9 років тому +1

    was of great help, i tried doing this on indian index i.e cnx nifty and got misclassification error of just 14 %

  • @arunblast
    @arunblast 8 років тому

    What is the accuracy value?

  • @jorgehenriquedefriasbarbos3253
    @jorgehenriquedefriasbarbos3253 7 років тому

    very good!
    tank you

  • @sagarkahane
    @sagarkahane 8 років тому

    Hi i have done same coding as u but missclasifi. error is 69 % how ? because yours is 51 % ? why it is diff. ? done all same as u.

  • @learn2earn167
    @learn2earn167 4 роки тому

    Warning message:
    glm.fit: fitted probabilities numerically 0 or 1 occurred