Post-anova item analysis Does the following corresponds to an existing process?
N subjects are subjected to m test items. 2 between-subject factors are manipulated with 2 modalities each. An ANOVA reveals that one of the factors is significant. 
Does it make sense to run an "item" analysis for the significant factor? This analysis would serve to determine the items (out of the m) this factor impacted the most? (or the items that contributed the most to the factor's significant effect).
I saw some papers doing so by conducting individual t-test for each item. I wonder, is this process is valid, and if so how is it called and/or are there alternative techniques?
Thanks for any advice!
 A: Yes, it's entirely valid to conduct a bunch of individual t-tests to determine which groups are different. However, when you run many tests, you increase Type 1 Error Rate - you may accidentally reject your $H_0$ when you shouldn't. 
The typical thing to do is to adjust your significance level $\alpha$. One way is to use Bonferroni Correction, which is 
$$\alpha^* = \alpha \cdot \cfrac{1}{K}$$
Where $K$ is the number of tests to run - in case of $k$ groups to compare it's $K= \cfrac{k \cdot (k - 1)}{2}$
Another way is to conduct a Tukey HSD Test, which essentially gives you the same information. 
Example in R:
Consider a data set of donuts with amounts of fat absorbed from donuts vs type of fat (taken from here)
t1 = c(164, 172, 168, 177, 156, 195)
t2 = c(178, 191, 197, 182, 185, 177)
t3 = c(175, 193, 178, 171, 163, 176)
t4 = c(155, 166, 149, 164, 170, 168)

val = c(t1, t2, t3, t4)
fac = gl(n=4, k=6, labels=c('type1', 'type2', 'type3', 'type4'))

boxplot(val ~ fac, col='skyblue')


aov1 = aov(val ~ fac)
summary(aov1)

Here the $p$ value is 0.00688, so we reject the $H_0$ and want to find out which groups are different
pairwise.t.test(val, fac, alternative='two.sided', p.adjust.method='bonferroni')
TukeyHSD(aov1)

Both tests say that type2 and type4 are different with $p$-value < 0.01
