Chi Square Expected Count less than 5!

[statshelpneeded]

Hi there, any help needed with my problem would be MUCH appreciated.

I have a 2x7 design. Gender VS different treatment options. I did a Pearson's Chi Square and my expected cell count was less than 5 in more than 20% of cases and some expected counts were also less than 1. SPSS does not allows for Fisher's Exact Test to be conducted on contingency tables larger than 2x2. Any ideas how i can get round this?

Thank you in advance!

 

[trinker]

You could bootstrap but I don't know if SPSS does this.

 

[gianmarco]

Hi,
You could:
1) collapse some categories in a sensible way,
2) use R

As for (2), the following links could help:
link
link
link

Hope this helps
Regards
Gm

 

[statshelpneeded]

I have seen bootstrapping on some SPSS tutorials, but i can't seem to find the option in my dialogue box. Not sure if it's an add on, will do some more research! Thanks!

 

[gianmarco]

You could bootstrap but I don't know if SPSS does this

Yes, as suggested by Trinker, I recall that there should be an option in the SPSS menu to tell SPSS to also perform bootstrapped chi-square.

Gm

 

[statshelpneeded]

Yes, as suggested by Trinker, I recall that there should be an option in the SPSS menu to tell SPSS to also perform bootstrapped chi-square.

Gm

Not sure if i'm doing it wrong, but no bootstrapping option? However, if i go on the exact tab I can select Monte Carlo, Exact or Asymptotic only. If i select exact, it specifies a time limit of 5 minutes per test and the output gives me a fisher's exact reading. Is this a solution?

 

[gianmarco]

Hi,
unfortunately, I cannot see the details of the pict. Besides, I have not my Mac at hand so I cannot check by myself. In any instance, I think that that solution could work for you.

Some info on Monte Carlo in chi-square context at this link

Regards
Gm