[Lisa Wilson]

Thanks
I am a junior researcher. I have created an index that captures investor sentiment using aggregate micro-level data. The index may have some influential observations and I want to ensure that they have no large effects on my results — I can not simply exclude them.

The ultimate objective is to test whether there is a significant association between stock returns and my sentiment index, using time series regressions. So far I used OLS regressions with Newey West standard errors and the results show a significant association between stock returns and the sentiment index. I replicated my time series regression using both -mmregress- and -robreg10 mm- and the inferences do not change. I understand that both are suitable for linear regressions in general but I am not sure whether there are any problems when applied to time series regressions.

Let me know if I need to provide any further clarification. Thank you

[Nick Cox]

In my experience "time series regression" can mean anything from "regression using time series" to "regression with a specific time series flavour" which could mean e.g. using lagged regressors, testing for serial autocorrelation, and so on, and so forth. This is the nub of the matter, perhaps, I think you now most need advice from fellow economists.

[Carlo Lazzaro]

Lisa:
three unsolicited advice:
1) be confident with your approach,do not ask for further support and face supervisor/colleagues/reviewers' opinion on your research as it comes (more or less, this is what we all do when we have to submit/deliver a research report, working paper, manuscript to be hopefully published after n-rounds of major/minor revisions. On a personal note, I am still celebrating the Xmas season this way, switching from author to reviewer and back again);
2) seek for help/advice from more eexperienced "time-serialists" among your colleagues;
3) be aware of the possible (too) partial specification of your regression model (it is frequent to find a significant predictor in a simple OLS. Unfortunately, it rarely means that the researcher has given a fair and true view of the data generating process she/he is investigating).

[Bruce Weaver]

Regarding time series quantile regression, this chapter (which I have not read!) might be helpful:

Xiao, Z. (2012). Time series quantile regressions. In Handbook of statistics (Vol. 30, pp. 213-257). Elsevier.

[Lisa Wilson]

In my experience "time series regression" can mean anything from "regression using time series" to "regression with a specific time series flavour" which could mean e.g. using lagged regressors, testing for serial autocorrelation, and so on, and so forth. This is the nub of the matter, perhaps, I think you now most need advice from fellow economists.

Thanks Nick. Yes I hope someone in economics/econometrics can also help!