Amir, yes you should use repeated measure analysis. But here with 9 measurements and 2 years with 2 categorical variables you should analyse in this way. Two way repeated measure ANOVA with years and 9 measurements as two way repeated. it should be noted to use two way repeated measure you need to have even measurements, 8 or 10. So remove or add another level to your 9 measurements. for example if you add a measurement then you have 2 years each 5 measurements. This is the two way repeated measures (2(one way)*5( another way)). So you can analyse these sources:

1- effect of year on data

2- effect of 5 or 4 (9-1)measurements on data

3- interaction between year and measurement times

4- interaction between categorical var with year

5- interaction between categorical var with measurement times

5- triple interactions.

Hope it helps

I agree with Amir Sariaslan. Repeated measures ANOVA assumes normality of residuals (among other somewhat strict assumptions) which is not going to happen with discrete outcomes. Mixed regression and latent variable models give you more options in the link (cumulative logistic, for example).

With 9 occasions, you also have a great deal of flexibility in the functional form of the trajectory. Most researchers in my field (psychology) focus on polynomial trajectories, but there are other options, such as exponential functions, which may be more appropriate depending on the hypothesized (and/or observed) shapes of the raw data trajectories.

Pat

While the factorial design is the same as in ANOVA, categorical variables need special statistical analyses. Repeated measurements are a form of two factorial design. General purpose analyses, like ANOVA or linear regression, are based on the normal distribution and can only be applied when certain conditions are met: 10-90% range, for percentages, or n>5 plus log transform, for counts. The appropriate methods can be logistic regression (for percentages), Poisson regression (for counts) or other chi-squared based tests. Small counts (lower than 5 individuals per cell) are a problem and may require the exact test approach. Box 17.16 in Sokal and Rohlf (2012), "Biometry", fourth edition, shows a worked example for 2 groups, 2 times. If you are a SAS user, there are other approaches in Stokes, Davis & Koch (2000), "Categorical Data using the SAS System". The theoretical basis can be found in Agresti (2013). "Categorical data analysis", fourth edition.