SpiderGuard 0 Posted April 1, 2008 Share Posted April 1, 2008 I have a non-normal distribution (normalish looking, but skewed right), and I need to figure out how to find z-scores. Computing z-scores is easy enough for a normal distribution, but I have no clue how to do it for non-normal distributions. Any help from the stats folks here would be greatly appreciated. Link to post Share on other sites
runthemover 39 Posted April 1, 2008 Share Posted April 1, 2008 look at the definition of what a z-score isedit: I see you said non-normal. I was wondering why this was so simple. uhh. not sure. I'll check my stat books Link to post Share on other sites
SpiderGuard 0 Posted April 1, 2008 Author Share Posted April 1, 2008 look at the definition of what a z-score isedit: I see you said non-normal. I was wondering why this was so simple. uhh. not sure. I'll check my stat booksThanks! I've poked through my stats books and haven't found anything...but I'm in the social sciences so we're a bit slower than the actual math types Link to post Share on other sites
runthemover 39 Posted April 1, 2008 Share Posted April 1, 2008 what is it exactly that you are given? what does the question ask? Link to post Share on other sites
SpiderGuard 0 Posted April 1, 2008 Author Share Posted April 1, 2008 what is it exactly that you are given? what does the question ask?It's an actual dataset, so I have pretty much any information you'd need or want So I have a mean, standard deviation (which if it were normal would be enough), and any other measures you'd ever want.Edit: The dataset has something like an N of 16000...not sure if there's any difference between t-score and z-score for non-normal distributions, but I thought I'd throw that out there. Link to post Share on other sites
runthemover 39 Posted April 1, 2008 Share Posted April 1, 2008 what does the question ask you to do?I'd say you should try to get the data to be normal by getting rid of any outliers, or separating by subgroups or something.It's hard to know what's going on without knowing the question. Link to post Share on other sites
ajs510 122 Posted April 1, 2008 Share Posted April 1, 2008 Thank you both for making me feel better about having dropped out of college before taking Statistics. My head hurts just reading that. Link to post Share on other sites
SpiderGuard 0 Posted April 1, 2008 Author Share Posted April 1, 2008 what does the question ask you to do?I'd say you should try to get the data to be normal by getting rid of any outliers, or separating by subgroups or something.It's hard to know what's going on without knowing the question.It's actual analysis on a paper, so there's no specific question. But let me give more detail on what I'm actually trying to do with this dataset.I have a set of distances between word-pairs (N=16600), and I'm trying to figure out how many z-scores a specific word pair is away from the mean of the distribution. I'm attaching the plot below - but basically what I'm trying to do is show that Americans and We are statistically further away from the mean than November is. Link to post Share on other sites
runthemover 39 Posted April 1, 2008 Share Posted April 1, 2008 uhh. weird. I'm kinda confused about what the heck you're doing but ok.I'd say you could probably just say that the distance between word pairs (ths is what the X axis is right?) for americans and we are in the 500 area whilst november has a distance of 900 or so and the mean is at 1050. I don't think you need the distribution to talk about that?Either that or send an email to your prof for clarification. I'm still a bit confused about what you're doing. Link to post Share on other sites
SpiderGuard 0 Posted April 1, 2008 Author Share Posted April 1, 2008 uhh. weird. I'm kinda confused about what the heck you're doing but ok.I'd say you could probably just say that the distance between word pairs (ths is what the X axis is right?) for americans and we are in the 500 area whilst november has a distance of 900 or so and the mean is at 1050. I don't think you need the distribution to talk about that?Either that or send an email to your prof for clarification. I'm still a bit confused about what you're doing.The whole word distance thing is kinda goofy - it has to do with priming and word relatedness. The prof delegated this down to me - I'm the methods guy on the team apparently, which is working out incredibly well for me We're looking for some sort of hypothesis testing framework in which we can discuss these numbers - which is why we want z-scores. Social scientists are used to z-scores and p-values, so that's what we're trying to give them. Something more "scientific"...good times Link to post Share on other sites
eYank 0 Posted April 1, 2008 Share Posted April 1, 2008 does it say anything like the population is normal? Id try computing it like its a normal distibution but im not sure... Link to post Share on other sites
Actuary 3 Posted April 1, 2008 Share Posted April 1, 2008 Z scores are not restricted to Normal DistributionsIf you need help calculating the Standard Dev or Mean, let me know. Link to post Share on other sites
SpiderGuard 0 Posted April 1, 2008 Author Share Posted April 1, 2008 Z scores are not restricted to Normal DistributionsIf you need help calculating the Standard Dev or Mean, let me know.So it's calculated in exactly the same way? If it's not normally distributed there are different percentages of the data on each side of the mean - doesn't a standard deviation mean more on the right side of the mean than the left? Link to post Share on other sites
Dawson Leery 12 Posted April 1, 2008 Share Posted April 1, 2008 look at the definition of what a z-score isedit: I see you said non-normal. I was wondering why this was so simple. uhh. not sure. I'll check my stat booksSo if you have the point where November falls you should be good to go. I'm not sure why you want it in z-scores and not can't just say it's -1.25 standard deviations away from the mean or something, but whatever I don't remember stats that well. Link to post Share on other sites
Actuary 3 Posted April 1, 2008 Share Posted April 1, 2008 So it's calculated in exactly the same way?Yes. Z = (X - Mean) / Stnd DevIf it's not normally distributed there are different percentages of the data on each side of the meanNot always true- doesn't a standard deviation mean more on the right side of the mean than the left?A Standard Dev is just a statistic to describe the distance from the mean, it's not related to left or right Link to post Share on other sites
Actuary 3 Posted April 1, 2008 Share Posted April 1, 2008 So if you have the point where November falls you should be good to go. I'm not sure why you want it in z-scores and not can't just say it's -1.25 standard deviations away from the mean or something, but whatever I don't remember stats that well.that's what Z score is.What Normal Z-scores allow is for you to convert it to a nice percentage Link to post Share on other sites
vbnautilus 48 Posted April 1, 2008 Share Posted April 1, 2008 I have to disagree with Actuary on this one. Calculating a z statistic is basically meaningless for a distribution with this shape. You should certainly not be making inferences about probability based on this Z score, as this distribution his highly skewed and probability values associated with Z will not apply. It's still not clear why you want a Z score. If you are trying to test for significance you will have to look into non-parametric statistics. Link to post Share on other sites
Actuary 3 Posted April 1, 2008 Share Posted April 1, 2008 I have to disagree with Actuary on this one. Calculating a z statistic is basically meaningless for a distribution with this shape. You should certainly not be making inferences about probability based on this Z score, as this distribution his highly skewed and probability values associated with Z will not apply. It's still not clear why you want a Z score. If you are trying to test for significance you will have to look into non-parametric statistics.How dare you!SpiderGuard says he wants a Z-score.I told him he can have a Z-score without a Normal DistributionWhat are you disagreeing with?"Probability Values Associated with Z-scores" ... seems like you are thinking in terms of Normal, as my post above your noted.ps. Why are you disreguarding the other Parametric distributions. Just curious. Seems like something else could be fit; but alas, I don't recall the general shapes of these. Link to post Share on other sites
custom36 5 Posted April 1, 2008 Share Posted April 1, 2008 I have a non-normal distribution (normalish looking, but skewed right), and I need to figure out how to find z-scores. Computing z-scores is easy enough for a normal distribution, but I have no clue how to do it for non-normal distributions. Any help from the stats folks here would be greatly appreciated.When I had to figure out z-scores, we had a table we were allowed to use. I'd look online. Link to post Share on other sites
Actuary 3 Posted April 1, 2008 Share Posted April 1, 2008 When I had to figure out z-scores, we had a table we were allowed to use. I'd look online.you are thinking Normal Dist, I'd bet. Link to post Share on other sites
custom36 5 Posted April 1, 2008 Share Posted April 1, 2008 you are thinking Normal Dist, I'd bet.Probably. I only took Psychological statistics. Link to post Share on other sites
speedz99 145 Posted April 1, 2008 Share Posted April 1, 2008 Probably. I only took Psychological statistics.$5 says Actuary just broke out into an extremely condescending laugh. Link to post Share on other sites
vbnautilus 48 Posted April 1, 2008 Share Posted April 1, 2008 SpiderGuard says he wants a Z-score.I told him he can have a Z-score without a Normal DistributionWhat are you disagreeing with?I've just never seen a z score applied to a non-normal distribution. I don't see what the purpose is. ps. Why are you disreguarding the other Parametric distributions. Just curious. Seems like something else could be fit; but alas, I don't recall the general shapes of these.It's standard practice that when you don't know the shape of the distribution you use a test that doesn't make assumptions about the underlying distribution of the data. Link to post Share on other sites
custom36 5 Posted April 1, 2008 Share Posted April 1, 2008 $5 says Actuary just broke out into an extremely condescending laugh.I don't think there's any doubt. Link to post Share on other sites
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now