"flat log likelihood encountered, cannot find uphill direction" when using GARCH command on index returns

[Mwai Mwanza]

I am using time series across 7 months focused on index returns on countries. I ran "arch Price Attack, arch(1)garch(1) het(Price Attack)" where 'Attack' is a dummy variable and Price is the price of my chosen index. I use the same index information; the time period over which the dummy variable equals 1 changes. I encounter the ""flat log likelihood encountered, cannot find uphill direction" error or the "could not calculate numerical derivatives discontinuous region with missing values encountered" error. How can I rectify this. If more detail is needed I can provide more information

[Carlo Lazzaro]

Mway:
welcome to this forum.
The warning message thrown by Stata means that there's some problem with the ML function.
Inspect your data and see if there's something strange and/or run your model on a shorter time span and see whether the warning message is thrown again.

[Mwai Mwanza]

Thank you Carlo this confirms my thoughts.
A small follow up: my data is pretty much exactly the same each time I run the computation, the only change is the number of dummy variables that equal 1 (centred around 31mar2012). Why would my data be okay for different lengths of time. I have attached the 4 versions of the data I'm using as I'm not sure if I'm being clear(identical apart from the Attack column).
Also when I run 'arch Price Attack, arch(1)garch(2) het(Attack)', I get results, what does this mean for me, why wouldn't the (1,1) variation work?