FOR MORE INFO GO THROUGH THIS VIDEO LINK https://youtu.be/Jsk6An-bPB8 When did we learn the theory on FX rate predictions? The theory of forward FX rates were discussed in week 2 to week 5. It is important to understand that a one year forward rate calculation using interest rates (IRP), for example, used the effective interest rate for the next year.   What is a time series model? The two models we discussed in week 6 are 1) moving average and 2) exponential smoothing. There was also seasonality and time trend that could be considered with time series data. Regression model is not a time series model.   What data should we collect? FX rates, cash rates, and inflation rates are the basic variables needed for the regression model. You must decide what the data frequency should be to match with the prediction model. Please review week 6 lecture notes and supplementary chapter regarding how one should select the data frequency and what are the disadvantages of the regression models. You can add any additional data as independent variables if you can explain why such IV would be useful in improving the predictability of the FX rates.   How do I run a regression? Regression was discussed during the week 6 workshops. The ppt slides and the dataset are available on Blackboard site.    Where do I get the forecasted macroeconomic variables to produce a forecast? The naïve way is to assume that the immediately following period’s values are the same as the current period. Another way is to use time series analysis for the macroeconomics variables and make a prediction. The final way is to gather information from reliable resources. For example. GDP growth expectations for Australia are announced by the World Bank, OECD, IMF, and RBA. Remember that they all have different values.        6. How do we make a prediction for 2015 while we have data till 2017? •Lets say you want to predict the FX rate for the end of Q1 2015 •Although we have all data till 2017 April, we assume that we only have data to the point where the estimation is made. (i.e. The end of Q4 2015) •So, we predict for Q1 2016 and then use the full dataset to evaluate the performance of the prediction.    Full Data     Prediction 1     Prediction 2   WON/NZD     WON/NZD     WON/NZD Mar-14 983.28   Mar-14 983.28   Mar-14 983.28 Jun-14 953.16   Jun-14 953.16   Jun-14 953.16 Sep-14 923.55   Sep-14 923.55   Sep-14 923.55 Dec-14 895.49   Dec-14 895.49   Dec-14 895.49 Mar-15 847.83   Mar-15 step 1) Predict   Mar-15 847.83 Jun-15 856.67   Assume you only have  data till end of 2014.   Jun-15 step 1) Predict Sep-15 830.97     Assume you only have  data till Q1 of 2015. Dec-15 856.79         Mar-16 875.69   step 2) evaluate by comparing  step1) with actual 847.83 in full data    Step 2) evaluate by comparing    step1) with actual 856.67 in full data Jun-16 855.36     Sep-16 837.92     Dec-16 873.97     Note: The assignment is designed in a way that suits a 3000 level course. This means that there are no direct questions such as 'use Absolute Error Method to calculate errors and comapred the errors to evaluate the model.' which makes the task too simple and easy for a 3000 level course. The purpose of the assignment is to challenge students to first know which methods exist for calculating errors and select the most appropriate one. This could make the assignment seem vague but it is essentially asking students to experience a similar decision making process as they will do after graduation. I do understand that a detailed assignment structure where the models are selected and all methods to be used are written in the assignment is easy to do, but remember that when you start working no one gives you that amount of details.   More FAQs 1.     Is regression model and single equation model same? Does the regression model belong to econometric model? or statistical analysis in technical analysis?  (if the regression model is belonging to statistical analysis, please let me know what part of week6 note and the supplementary chapter for week6 deal with that.) Answer: Single equation model is one of many ways a regression model can be run since it limits the number of independent variables to just one single variable. Regression model can be both an econometric or technical model. If the independent variables are economic factors then it is an econometric model but if the independent variables are past FX rates then it is a technical model. 2.     How can I figure out coefficients? In fact, I also confuse how to set up dependent variable and independent variables.  (As you said in FAQs, because FX rate, cash rat and inflation rate are basic variables, I will set these three as independent variables. And I think spot rate is dependent variable. Please comments about this, too.) Please carefully read through the supplementary slides provided in week 6 linked to regression models. The slide will show you how the variables should be set up and how to interpret the results. 3.     Did only the single equation model (maybe regression model?), time series models, autoregressive model learn in workshop? By the way, Is autoregressive model one of category in time series models? We also did a multiple equation model. Autoregressive model in one of the time series model but is based on technical analysis since only the past FX rates are used. All models in the lecture notes were discussed or are a small variation which students can try. 4.     If you have any data, book, link, youtube video and so on about regression model and time series models, please give them to me.   Try these links which are used in my other course. For more videos try the recommended links that show up in the youtube clips provided. https://www.youtube.com/watch?v=l3fxRVr1GRo https://www.youtube.com/watch?v=vzreSCd0bM8