I am trying to interpret some KPSS results for a large sample (n=1800 - http://data.is/TB2z5b). In general the KPSS test can be used for samples with time series with more than 100 observations? Is there a reference I can consult?
The KPSS test will work with large samples. However, it will almost always be significant, as with all hypothesis tests, when applied to large samples. The null hypothesis of stationarity is only really true in exceptional circumstances (e.g., when the data are simulated). When there is enough data, there is sufficient evidence that the null hypothesis is incorrect.
-
1$\begingroup$ I searched on google scholar and i was not able to find any good reference on this issue, could you suggest a paper i could consult ? In the case of large samples how could someone address this limitation ? $\endgroup$ – nkorf Nov 26 '12 at 15:08