iSixSigma

Correct use and analysis of ANOVA?

Six Sigma – iSixSigma Forums General Forums General Correct use and analysis of ANOVA?

Viewing 21 posts - 1 through 21 (of 21 total)
  • Author
    Posts
  • #53464

    Rider
    Participant

    I see some of the forum gurus are back on. I am prepared for the assault, but I’ll pose my query anyway:
    I have performance rankings for several hundred employees at a particular location. I have a group of current employees and a group of terminated employees and their rankings. I ran an ANOVA and obtained F=0.516 at F-crit=3.9. My null is that there is no difference in the rankings of employees who stay and those who quit voluntarily. With my results I think I should fail to reject the null. I have a VP, however, who is emphatic that he knows for certain that those who quit are always low performers. My results don’t seem to support that assertion. Is ANOVA a proper tool in this case or am I just off the mark again?

    0
    #190220

    Kluttz
    Member

    Define “employee rankings”. Is it a 1-5 scale? A straight ranking 1 through n?

    0
    #190221

    bbus
    Participant

    There are many interesting points

    1. What is a VP ?
    2. In my exp . only the best people leave first ( In a sinking ship , the rats ( best swimmers) are the first to jump !
    3. You are dealing with ordinal data ( ratings ) – do yu think ANOVA is the right tool ? Without ensuring that , you cannot make conclusions

    Just my opinion

    0
    #190222

    Rider
    Participant

    Employees are ranked from 1 (the top) to n (the bottom).

    0
    #190223

    Rider
    Participant

    1. Vice President
    2. Even a cursory look at my data does not indicate which rats are hitting the water.
    3. My point exactly, which is why I’m asking.

    0
    #190224

    Kluttz
    Member

    Mann Whitney

    500 dollars please.

    0
    #190225

    Rider
    Participant

    Hmm. Ordinal data, independent samples of dissimilar size. I really need a mentor.
    Thanks.

    0
    #190228

    MBBinWI
    Participant

    Try binary logistic regression. Your two states are employed, or quit. Your predictor is rank. If you want to include employed, quit, or fired, you could do a nominal logistic regression.

    Good to see you on the board, Ghost.

    0
    #190240

    Rider
    Participant

    Okay, now we’re getting over my head.

    0
    #190241

    the dude
    Member

    What action to you or your boss propose to take on your findings?

    0
    #190244

    Rider
    Participant

    We are attempting to see if we can predict the likelihood of someone quitting employment by looking at their performance ranking in the company. I have 1332 data points; some are still employed, others not. I’ve run the regression test mentioned above, but it’s way over my head and analyzing the results is even more over my head.

    0
    #190246

    bbusa
    Participant

    why do yu take up tasks which are beyond you ?

    0
    #190258

    Rider
    Participant

    Thank you for the question bbusa. Though it really is not relevant to this thread, I’m always up for a bit of banter on the forum. Before I respond, however, I would like to send thanks to MBBinWI, who provided information that has allowed me to find several of the missing pieces to my puzzle. Such relevant, germane, and fitting (proving a point with synonyms here) advice are the hallmark of collaboration, in my opinion.

    Now back to your question. I can only infer that you have seen other of my postings in the past as your question denotes a bit of historically present sarcasm in it; fair enough if so, my apologies if not. If you have seen past queries then you already have the answer to your question and I wouldn’t insult you by regurgitating that which you must know.

    If you have not seen my past posts, then your question is just a bit of banter, or you’re just a jerk. I’m cool with it either way. I won’t even mention that you seem ill-equipped to write standard English-you should learn how to begin a sentence or question as this concept seems beyond you. Moreover, I would suggest reviewing any English grammar book you may have handy and learn the proper spelling of the second person singular; spelling “you” correctly seems beyond you. The whole “which” and “that” thing in American English is probably definitely beyond you so I won’t even get into it.

    Now that I’ve had my turn at banter, the answer to your question is “because I’ve been told to.”

    0
    #190264

    MBBinWI
    Participant

    Ghost: My pleasure. Hope the “task that was beyond you” has been accomplished.

    0
    #190267

    bbusa
    Participant

    “I’ve run the regression test mentioned above, but it’s way over my head and analyzing the results is even more over my head.”

    Why the hell did you run a reg. if the analysis was way over your head?

    BTW : Your English is funny , too . You don’t need to speak Queen’s English to understand regression !

    0
    #190268

    Venugopal G
    Participant

    Ghost Rider wrote:

    I see some of the forum gurus are back on. I am prepared for the assault, but I’ll pose my query anyway:
    I have performance rankings for several hundred employees at a particular location. I have a group of current employees and a group of terminated employees and their rankings. I ran an ANOVA and obtained F=0.516 at F-crit=3.9. My null is that there is no difference in the rankings of employees who stay and those who quit voluntarily. With my results I think I should fail to reject the null. I have a VP, however, who is emphatic that he knows for certain that those who quit are always low performers. My results don’t seem to support that assertion. Is ANOVA a proper tool in this case or am I just off the mark again?

    Hi,

    As per my understanding, Test conditions are as below

    1. Compare two groups – Stay Vs Quit
    2. Data Type – Ordinal and it is Attribute data

    So ANOVA cannot be used. Since the data is Attribue 2-sample poisson rate test should be used.

    But I dont think so that we should consider rankings as continuous.
    But there are some people who do consider them as continuous. If so the tools are as below

    1. If they are indendepent samples, and focus is to compare mean or median and when the data is normal, use 2-sample t test.
    2. If they are indendepent samples, and focus is to compare mean or median and when the data is non-normal, use Mann-Whitney test.
    3. If they are indendepent samples, and focus is to compare standard deviation or variance , use 2 variances test.
    4. If they are dendepent samples, use paired t test.

    But one basic question my friend. You are talking about predicting the likelihood of someone quitting employment by looking at their performance ranking in the company.
    But have you done the MSA on the data collected? How can you prove the data collected is calibrated and we can rely on the same? Because the whole prediction is based on how correct the data is. According to me you will not get a good data i.e., employees giving honest ratings which when analysed says that these employees are planning to leave or stay back. Just a thought.

    Venu.

    0
    #190275

    Rider
    Participant

    If bbusa is trying to achieve something with what appears to be a dressing down of yours truly, I fear he has failed. You see, bbusa, your last question only provides the evidence that I didn’t actually need to show that you have no clue what you are talking about. Knowing something and being able to clearly explain it are two different things. But if it makes you feel better then go ahead. And understanding reg is great, but what if you can’t explain it?

    0
    #190276

    bbus
    Participant

    Jerk or no Jerk , I simply can’t get over the fact the you tried regression for the situation described & expected others to show you the way !

    You owe an explanation to your organization that must have spent a few thousand dollars to get you trained .
    BTW : I’d like to be paid to explain regression in a way that you would understand

    0
    #190280

    Argueta
    Member

    I would use chi-squared test.

    0
    #190380

    Rider
    Participant

    Hi bbusa,

    It’s been quite some time since I’ve posted as I decided to take your advice. Sort of. Before I explain, however, I must digress. I did not expect anyone to show me anything. This forum is meant as a means to guide those of us who are clueless; to a point. My Wisconsin friend indeed guided me, you did not. Also, you make assertions that a true Black Belt would not; my company has spent no money, zero, zilch, nada on getting me trained. That is probably part of the problem. Someone got six sigma in his head and I got picked. I’m attempting to learn this all on my own, with the help of course of some folks I’ve met recently. Also, I would not pay you for anything. It took all of about 20 minutes for someone to explain regression to me. Your services weren’t needed.

    You showed your hand in a post when you mentioned the Queen’s English. It’s referred to as the King’s English as a result of the King James Bible. It’s simply a figure of speech that you bastardized in a failed attempt to show your intelligence. We see what that got you.

    Anyway, I’m done with my light-hearted banter, pal. I went on a forum here in Houston and decided to see if there were any Six Sigma groups around. There were. Guess what; I went. When the head guy asked if anyone had questions, I piped up. I explained my situation and got the help I needed.

    Those who might recognize me on this forum know I’m the first one to proclaim my ignorance. You, conversely, are too stupid to do so.

    Later loser, and don’t respond. I’m done waiting for the doc to sand my bunions, and you aren’t important enough to waste otherwise good time.

    0
    #190385

    MBBinWI
    Participant

    Ghost: Wasn’t worth the time. I just ignore ignorance. Glad to hear you found a local support group. You can always post here and wade through the inane banter. You at least are diligent enough to do as much as you can before asking for help – more than I can say for a number of BB’s (that I’m working with, and that come to this forum).

    0
Viewing 21 posts - 1 through 21 (of 21 total)

You must be logged in to reply to this topic.