I need help. I've to use stata to estimate some consumption stuff in an economics class and I haven't a clue how to do it. This is long and it took me ages to type lol. I've done some of the work I think but I need some guidance... RapidShare: Easy Filehosting
That's the project. I've done all the stuff it walks you through in the project pack.
Let me say that I'm not looking for anyone to just do my project (But I won't be mad if you do), but really I just need some kind of general hints and tips.
Here's the dataset. RapidShare: Easy Filehosting Its Navidata stuff. I've uploaded it in the required edited form from Excel. You import in to Stata and check the comma delimited thingy.....maybe you people already know that. The instructions are in the first upload anyhow.
To start, I put the CSV file into the working directory and import all of the data:
insheet using NAVIDATA_Excel_Edited.csv, clear
Per the instructions, I've relabeled the variables with the following commands:
label variable y "consumers' expenditure (£ million 1990 prices and defined in CSO, CCBH)
label variable x2 "personal disposable income (£ million 1990 prices, CECO)
label variable x3 "interest rates (Treasury Bill yield %, AJRP)
And applied the transformation to y and x2:
I've done some regressions too:
regress y x2 in 1/21
regress y x3 in 1/21
regress y x2 x3 in 1/21
I should be doing all my regressions in the form
because it asks data for <1993 for questions 2-3 shouldn't I?
Right the ttests. Say we do the
regress y x2 in 1/22
function. Do I use the tvalue for x2 or the _cons? What is the _cons figure anyway?
Say we look at the x2 tvalue which is 35.86 and the pvalue is 0.00. This would imply a linear relationship as 35.86 is quite significant from the .897 and 1.009 95% confidence intervals. Is that right?
y = a1 +b1x2 + u => MPC => dy/dx2 = b1
Apparently the hypothesis test is just as given for b1? What does that mean? Like test at 95% b1 = 0?
To test the significance of a1,a2 b1,b2 do I work them out and test them manually?
And is it the same for finding b3's confidence intervals and b1,b2,b3's significance in Q.3?
To add a time trend I did this:
generate float t = year-1971
and this to generate a lag variable:
generate float ylag = y[_n-1]
which is fine I think.......?
To forecast the consumption for 1993-1996 I think I use corrgram command in some way but how does that tell me consumption levels? I tried
corrgram y, lags(24)
but I don't really know what that did......
Then I tried this:
twoway connected y year if year>1992
and I got a graph with consumption and years on the axes....would this be a correct forecast? I think it might be ?
Or do I use the arima and tsline commands somehow?
What does "log linear form" mean? It's Q5....
Man I can't tell you how much I appreciate your help guys. Thanks loads.