questions about CHI-squares...

[alleyez]

---

Members don't see this ad.

so Ho is basically observed = expected and Ha is the opposite.

i need a quick rundown of what goes in the discussion section of a lab report dealing with chi-squares.


I talk about my p value and how sinceit was less than 0.5, I will accept the null hypothesis instead of considering the alternative hypothesis...

what else do I include? I only have half a page and not sure what else to include...

thanks

 

[exmike]

Originally posted by alleyez
so Ho is basically observed = expected and Ha is the opposite.

i need a quick rundown of what goes in the discussion section of a lab report dealing with chi-squares.


I talk about my p value and how sinceit was less than 0.5, I will accept the null hypothesis instead of considering the alternative hypothesis...

what else do I include? I only have half a page and not sure what else to include...

thanks

you mean P is greater than 0.05

 

[alleyez]

Originally posted by exmike
you mean P is greater than 0.05

yea my bad the P was less than 0.05 not 0.5

so I accept the null hypothesis of the observed equalling the expected. right?



Also, how do I figure out my P for this other data set if x^2 was 140 with the degree of freedom of 3?


How do look that up?

I just want this thing over with.. sorry for the questions. Hopeuflly someone can answer some of these questions

 

[exmike]

i'm confused. i took biostats two years ago so maybe i'm fuzzy. The null hypothesis is rejected if the difference between two groups isnt statistically significant. So you're accepting it meaning it IS significant. Significant is the P< 0.05.

So which is it?

 

[alleyez]

Originally posted by exmike
i'm confused. i took biostats two years ago so maybe i'm fuzzy. The null hypothesis is rejected if the difference between two groups isnt statistically significant. So you're accepting it meaning it IS significant. Significant is the P< 0.05.

So which is it?

my P is

0.05 < P < 0.025

so it is significant right?

it seems to make sense....


I also have some questions in the original post... can someone please give me some guidance.

 

[spaz]

Originally posted by exmike
i'm confused. i took biostats two years ago so maybe i'm fuzzy. The null hypothesis is rejected if the difference between two groups isnt statistically significant. So you're accepting it meaning it IS significant. Significant is the P< 0.05.

So which is it?

usually the null hypothesis proposes no difference. so you reject the hypothesis if there is a statistical difference.

 

[exmike]

you put 0.5 in your original post

 

[alleyez]

Originally posted by exmike
you put 0.5 in your original post

yea sorry about that, i meant 0.05

 

[spaz]

if your p-value is < 0.05, you must reject your null hypothesis, because the difference is statistically significant.
the p-value represents the probability that the difference was due to normal variation. usually we use 0.05 as the benchmark for determinining significant vs. insignificant.

sorry don't remember too much about chi-square tables in particular. but there are tables that you use to compare your chi-square value to the critical value associated with the # of degrees of freedom

 

[spaz]

here's a link to website i used as a reference when i took biostat... hopefully it'll help

http://davidmlane.com/hyperstat/chi_square.html

 

[spaz]

these forums always seem to be so slow to load for some reason.


anyway, here's a link to a chi square table
http://www.ento.vt.edu/~sharov/PopEcol/tables/chisq.html

w/ 3 degrees of freedom, a chi-square value of 140 is definitely much greater than a chi-square of 16.27 ( p=0.001 ). although 140 isn't on the table, we know its p-value would be <<<< 0.001. therefore your results are statistically significant.


ok enough learning (er, re-learning) for me.

 

[MD08]

Hey-

You mentioned that you needed material to put in the discussion section... I am not sure how advanced you are in your statistical capabilities, but I would mention that the chi-square statistic only tells you whether or not there is an association between your two variables, and not the magnitude of the association. Also, it will not tell you if the association between your two variables is different depending on which categories you are looking at.

Of note, the chi-square statistic can be sensitive to how many categories you have split your data into. However, since your result is highly significant (p<.001), I doubt this would make any difference.

Another thing to check- Do any of your cells in your table have an expected count<5? If so, the chi-square statistic is not valid, and you would need to use fisher's exact. Also, are your observations independent? This is another assumption of the chi-square stat.


Good luck.

 

[alleyez]

thanks for all the posts so far guys... it has helped.



If a p value is statistically significant.. is the Ho ALWAYS rejected? or even though the p value might be statistically significant, you decide to keep the null hypothesis?

under what circumstances can i still accept the null hypothesis? how do i explain my reasoning

 

[MD08]

Hey-

You mentioned that you needed material to put in the discussion section... I am not sure how advanced you are in your statistical capabilities, but I would mention that the chi-square statistic only tells you whether or not there is an association between your two variables, and not the magnitude of the association. Also, it will not tell you if the association between your two variables is different depending on which categories you are looking at.

Of note, the chi-square statistic can be sensitive to how many categories you have split your data into. However, since your result is highly significant (p<.001), I doubt this would make any difference.

Another thing to check- Do any of the cells in your table have an expected (not observed) count<5? If so, the chi-square statistic is not valid, and you would need to use fisher's exact. Also, are your observations independent? This is another assumption of the chi-square stat.


Good luck.

 

[MD08]

Oh, to answer your other question:

Usually, p=.05 is the gold standard for rejecting the null, and if you observe this, you would not conclude anything otherwise. If you observe a p=.06-.07, you may mention that your result is of "borderline" significance. Although rare, some studies may choose a type-I error rate (p-value) of .10 before you reject, but this is pre-determined before the analysis or study is performed.

 

[alleyez]

what I don't understand is.... my P value is lower than 0.05

however, it doesn't make sense to reject the null hypothesis since my data of the observed is very similar to the expected values.

Can someone please explain this?
thanks

 

[Mediculous]

Try redoing the chi-square test if you think that the observed numbers were very proximal to your expected numbers. You said that your value was 140, but if the observed-expected numbers were similar, the test would be taking the sum of numbers very close to zero and thus your chi-square value will be low, not high, which 140 is. 14 is pretty high. 1.4 is a value you can feel comfortable with. Double check your math, and if you still get a very large chi-square value, and p<.05, you have to start thinking about alternative hypotheses. Like I mentioned before, I think that in a general biology fruit fly lab, especially the first one, the profs. have the lab designed to make you reject your null hypothesis, which is normally a simple genetic system, and make you think about alternatives, just so you get accustomed to using the scientific method. Anyhow, double check your calculations and let us know what's going on.

 

[alleyez]

Originally posted by Mediculous
Try redoing the chi-square test if you think that the observed numbers were very proximal to your expected numbers. You said that your value was 140, but if the observed-expected numbers were similar, the test would be taking the sum of numbers very close to zero and thus your chi-square value will be low, not high, which 140 is. 14 is pretty high. 1.4 is a value you can feel comfortable with. Double check your math, and if you still get a very large chi-square value, and p<.05, you have to start thinking about alternative hypotheses. Like I mentioned before, I think that in a general biology fruit fly lab, especially the first one, the profs. have the lab designed to make you reject your null hypothesis, which is normally a simple genetic system, and make you think about alternatives, just so you get accustomed to using the scientific method. Anyhow, double check your calculations and let us know what's going on.

well... there are 2 model crosses not just one.

in the first one, the null hypothesis makes sense.... except the p value is <0.05. So i dunno if the prof. wants me to reject it or do something else with it.



in the second one... the null hypothesis is definitely rejected.

 

[Mediculous]

Huh, I don't really know what's going on. For the cross that you think supports your null hypothesis, the p is going to have to be >.05. If the p<.05, that means that the chances of your results randomly occuring within your hypothesis<.05, so you're saying that the chance of your results occurring, under your null hypothesis, are very low---which doesn't support your hypothesis. Was there a certain set of data that threw your value off? If so, you could mention that although the chi-square test suggests the null hypothesis is false, if not for a certain set of data, the chi-square test would support the null hypothesis.

 

[Bones2008]

Originally posted by alleyez
If a p value is statistically significant.. is the Ho ALWAYS rejected?

usually, but if cost < $50/hr, you can probably keep her around.