Regarding alternatives of t test when time series data is not normal I want to test whether a investment strategy generates significant return or not. For this I have to apply t.test. But my time series data is not normal. It is negatively skewed and have heavy tails. Please recommend me which test I should use.
 file shared tsmom
 A: I think your data suffers from auto-correlation and hetroskedasticity. To counter this issue in time series data mostly Newey-West t test is used. 
A: Data like yours can arise from a model that might include seasonal ARIMA structure or seasonal dummies ... the eye can't easily sort this out: Only the data knows for sure!
I took your first (of 12) time series ...192 months over 16 years 

and found

Your data appears to potentially have no significant autoregressive seasonality. It is another question as to whether or not there is deterministic seasonality i.e. the need for monthly dummies.
I then investigated the applicability of seasonal dummies and found that there indeed was the suggestion of deterministic seasonality and an ar(2) component unmasked after adjusting for "anomalies" which left untreated can obfuscate the model.

The residuals from this model are here 
with companion acf here

In terms of the next step ... I would estimate the seasonal dummy model individually for each of the 12 AND then estimate the same model for the composite of the 12. An F test much akin to the CHOW TEST can then be conducted to test the hypothesis that there is a common DGF ( Data Generating Function) and then proceed to efficiently segment the 12 into homogeneous classes.
The great thing about actually doing statistical analysis on unseen data is that it can change what we think we know to be true.
In that spirit , I suggest a modification to @Glen_b 's reflections .
"Two big problems are going to be possible autocorrelation" to the more encompassing 
"Two big problems are going to possible predictability within the time series" 
In closing , I have always been in awe about the positive attitude that @Glen_b brings to the table as he teaches and learns about statistical analysis providing an example of how moderators should attempt to modify their theoretical knowledge of statistics with significant empirical "data understanding" suggests that.
