For the statisticians of the ORG

[FactionHeir]

Hello :smiley2:

I've been running some chi square tests in an excel sheet where I have substituted some old baseline data with newly gathered data.
The problem is, there was just about no documentation in that file other than being set up in a way that its relatively simple to know where to put what data.
That said, the tables are setup to spew out some chi square values, but I don't really understand what the numbers mean and whether they really are significant or not.
Also, I don't know if the df is correct and what kind of p value (significance) the data has.

Sooo...if you feel like helping out and know something about chi square and data analysis, I'd appreciate if you could have a look at the following excel file I uploaded (1 sheet only):

http://download.yousendit.com/63ED2FCA707B9F14

I marked my preliminary thoughts in red.

Thanks for helping :2thumbsup:

[econ21]

Well, I confess I have never been taught chi-squared tests, so take the following with a pinch of salt. But digging around a little, there are three excel chi-squared functions that can help you.

Chitest() can be used to produce the p-value from the raw data on observed and expected values, as explained here:

http://udel.edu/~mcdonald/statchigof.html

For example, I can take your cells B4:B24 and stack them with C4:C24 to get a column of observed values. I can then make a column from D4:D24 and E4:E24 to get the expected values. Putting these two columns in to the chitest() function plops out the figure 0.9655 for the p-value. Wholly insignificant => we can't reject the 3:2 ratio.

If you do want to do the t-testing more manually, you might like to take the chi-squared figure you compute, say the 21.04 in cell M23 and use the chidist() function to get the p-value associated with your statistic. Degrees of freedom for an extrinsic null hypothesis are (number of observations -1), so that makes 35. If we compute chidist(21.04,35)=0.9655 which is what the chitest function told us already.

If you want to report the critical value, you can use the chiinv() function. For example, if we want to know it at the 5% level, we use chiinv(0.05,35) and find the critical value is 49.80. So our statistic (21.04) is well below that.

I think the above should be enough to get you going.

If you want to know more about what is going on, this is a link that talks you through a similar kind of null hypothesis to your sex ratio ones (a 50:50 split):

http://egret.psychol.cam.ac.uk/psychology/2003-4/1B_Stats_Handout_5_chi-squared.pdf



EDIT: A worthy post #7000 for me, I think. I just hope some real statistician does not shoot the above down!