Long Shot but....does anyone know anything about...
sophie78
Posted:
Member since
Long Shot but....does anyone know anything about...
How to test whether data is linear or not before using a Pearson's Correlation? Pearson's can only be used if it is linear. Do I make a scatterplot? So confused! Anyone have any ideas?
Janice
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
all greek to me
sophie78
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
You teased me! I thought I had my answer!!!
The Original 2nd-time-around
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Posted by am470
How to test whether data is linear or not before using a Pearson's Correlation? Pearson's can only be used if it is linear. Do I make a scatterplot? So confused! Anyone have any ideas?
A scatter-what???? Not a clue
Janice
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
sorry, I feel guilty when I read something and don't post. I hate it when posts don't get at least one answer.
RedHead
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Posted by am470
How to test whether data is linear or not before using a Pearson's Correlation? Pearson's can only be used if it is linear. Do I make a scatterplot? So confused! Anyone have any ideas?
in grad school i used only the Pearson... never did scatterplots
sophie78
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Posted by RedHead
Posted by am470
How to test whether data is linear or not before using a Pearson's Correlation? Pearson's can only be used if it is linear. Do I make a scatterplot? So confused! Anyone have any ideas?
in grad school i used only the Pearson... never did scatterplots
See you can't do Pearson's unless you test the data for linearity. But I totally forgot whether you eyeball it or whether there is a test for it. I hate dissertations!! And I hate professors who vacay all summer!!
RedHead
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Posted by am470
Posted by RedHead
Posted by am470
How to test whether data is linear or not before using a Pearson's Correlation? Pearson's can only be used if it is linear. Do I make a scatterplot? So confused! Anyone have any ideas?
in grad school i used only the Pearson... never did scatterplots
See you can't do Pearson's unless you test the data for linearity. But I totally forgot whether you eyeball it or whether there is a test for it. I hate dissertations!! And I hate professors who vacay all summer!!
sorry...i didn't have to do that.
cluelessbride
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Um, this is my really dumb answer but can you plot and if the slope is greater than 1 it's linear? Is it the plotting part that you need help with?
sophie78
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Posted by cluelessbride
Um, this is my really dumb answer but can you plot and if the slope is greater than 1 it's linear? Is it the plotting part that you need help with?
No that's not it. But thanks anyway.
Preshy7
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Re: Long Shot but....does anyone know anything about...
The scatterplot would show how strong the correlation is, and let you know whether run a Pearson's Correlation. If you can tell there is not a strong linear relationship by eyeballing it during the scatterplot, then it might not be worth it figuring out the Pearson's Correlation. We are using SPSS for our dissertations--do you use that? When you plus in the numbers it runs pretty much all of the tests, and shows scatterplots, etc.
MCDO15
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
lol, i just read the same thing...i think im a expert now...
i believe you want to do a scatterplot first to determine if its liner or not..then you can proceed
dpli
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Most data is never completely linear but may have a linear relationship. From what I read about the Pearson's correlation, you are checking for linearity, so you should be able to do that with a scatterplot. I am not a behavioral scientist, though, so maybe it is used differently in your field.
Edited, b/c I think my first answer was wrong, sorry.
sophie78
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Thanks but that is too simplistic for what I am doing.
sophie78
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
Posted by nov04LIbride
The scatterplot would show how strong the correlation is, and let you know whether run a Pearson's Correlation. If you can tell there is not a strong linear relationship by eyeballing it during the scatterplot, then it might not be worth it figuring out the Pearson's Correlation. We are using SPSS for our dissertations--do you use that? When you plus in the numbers it runs pretty much all of the tests, and shows scatterplots, etc.
I am using SPSS. Most of my correlations are insignificant. The scatterplots seem to show no correlation but the data seems to be linear and not curvilienear. So eyeballing it is the answer! I also consulted my textbook which said that most social science data is linear. So I think my data is linear but not correlated for the most part!
Thanks. I hate dissertations.
sophie78
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
I am using SPSS. Most of my correlations are insignificant. The scatterplots seem to show no correlation but the data seems to be linear and not curvilienear. So eyeballing it is the answer! I also consulted my textbook which said that most social science data is linear. So I think my data is linear but not correlated for the most part! So I can use the Pearson's!
Thanks. I hate dissertations.
dpli
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
My other thought is: can you use the variance values of your data to determine the linearity? Like if the variance is over a certain # you cannot consider it linear? If so, you could do a spreadsheet to calculate the variance values if your data is already entered as if you were doing a scatterplot pretty easily.
ETA: Never mind - it looks like you figured it out
sophie78
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
dpli-I don't think it works that way. But I could be wrong. The Levene's test for equality of variances when I did my t-tests all came up non significant which is good.
sjm71505
Posted:
Member since
Re: Long Shot but....does anyone know anything about...
I am getting my doctorate in psych too...so let's see if all those classes in stat paid off!
I would say that the correlation is the 'test' for linearity. If the correlation is signficant that means there is a significant relationship between the variables. If there is a curvilinear relationship, the correlation may come out to be insignficant because a product-moment corelation coefficient is sensitive to a LINEAR relationship between two variables . However, just because the P-Moment Coeff. is insignificant does not mean there is NO relationship between the variables. If you do a scatter plot, the curvilinear relationship may become more evident, which would then make your insignficant finding significant (in a nonstatistical sense). there may be some other test or analyses that can be done to see if the curvilinear relationship is signficant....but I don't know what it is off hand. HTH Now, if only I could get motivated to work on my dissertation
