Point me in the right direction?

#1
Hi all, this is my first post here, so if I'm out of line or in the wrong spot, just let me know.

Anyway, I've got a problem I could use some help with
Some background:

Test subjects are mice with determined genotypes, injected with cancercells and monitored for one year.
SIRT1 - promotes longevity in cells
p53 - tumour supressor (anti-cancer effects)

Ageing promotes upregulation of SIRT1 which consequently increases cancer risk by dacetylization anf inactivation of p53

Data set:
SIRT1....p53......Alive......CancerMort.....Noncancer Mort.....TotalMort
+/+......+/+......15...............8....................9...................17
+/+.......-/+......33.............17....................18...................35
+/+.......-/-.......13.............13.....................8...................21
+/-.......+/+..... 26.............11.....................9...................20
+/-.......+/-...... 51.............19....................17..................36
+/-........-/-.......22.............15.................... 7...................22
-/-........+/+......37..............6......................6...................12
-/-.........+/-......78..............8......................7...................15
-/-..........-/-......14..............26................... 6................... 32


the SIRT1 and p53 columns are the genotypes,
Alive = # of subjects alive after one year
Cancermort = # of subjects died of cancer after one year
Noncancermort= self explanatory
TotalMort= again self explanatory

Now i want to test the hypotheses that
a) SIRT1 inactivation reduces longevity
b)p53 inactivation increases cancer risk
c) that the effect of SIRT1 activation on cancer risk is due only to the consequent inactivation of p53

Now, is this a contingency table? I'm pretty confused and am really just looking for a place to start, and suggestions about what type of tests/analysis I should try. I don't know why I'm having such a hard time with this, it seems like it shouldn't be that difficult. I am using SPLUS 7.0.

Thanks alot for any help.:tup:
 
#2
ummm, I know its only been a couple days, but I am concerned about this getting knocked down and off the main page. If nobody has any opinions on this, thats cool, but I would REALLY appreciate some input. Thanks.
 
#3
So.........I am assuming that I need to look at this as a multinomial GOF scenario: where the expected values are that the proportion of survivors will be higher in SIRT1 + mice, looking only at the Alive/Total.Mort columns? And the cancer.mort should be a higher proportion of the total.mort when p53- mice are looked at? How do I account for the hypothesized interaction of the two factors when testing for the effect of each indepedently? Still confused...
 
#4
a) It appears that you are not interested in cancer for this question, only whether the animal is dead or alive. I think a 3X2 contingency table would work here. Where the rows are the SIRT1 genotype and the columns are "alive" and "total mort". A chi-square test for homogeneity will work to test for a difference in the proportion of alive animals between genotypes. Since genotype is an ordinal type category, you might want to do a trend test to see if there is a downward trend in this proportion across genotypes.

b) You might do something similar here, where colums are "cancer mort" and "non-cancer mort".

You said:

How do I account for the hypothesized interaction of the two factors when testing for the effect of each indepedently?

This is an interesting question, and the above analyses do not account for this. I'll have to think about this, maybe dig out an old book. I'll try to get back to you.

I imagine that you are conducting this research in a university setting? If so, perhaps you should consult with a statistician from your math dept.

~Matt
 
#5
Thanks, I am going ahead with what you suggested in terms of the first two hypotheses, and see what I end up with. As for the third, I'm still a little shady on the format for testing for the independence.

Thanks alot Matt, I really appreciate you taking a look at this.