cross section fixed, period fixed or both

[barbaros_y]

hi,

I am working with 215 cross sections and 6 periods. when I do yearly regressions, for the first three years I find positive and significant relationship. in the last three years I find positive but not significant relationship.
when I run a pooled least squares I find positive but not significant relationship. then I do a hausman test and it points to fixed effects. but nevertheless there is a negative an significant relationship.

then I run cross section fixed effects model and I find a negative and signifacant relationship.

then I run a period fixed effects model and I find a positive and significant relationship.

I run both period and cross section fixed effects and I find a negative relationship.

which is my model: how should I select a model that explains the relationship between my dependent and independent variables? is it negative or positive.

by the way I divide the data into two periods: first three year and last three years. and I run cross section fixed effects model. first three years there is positive but insingificant relationship. last three years there is negative relationship.
it seems to me that the relationship between the dependent and independend variable changed.

also I want to ask about diagnostics?
what should I do
first unit root
chow for what
breusch pagan
others???

and how

thank you in advance