Hello; can anybody help me? I have to compute a exponential moving average of the S&P500 but when I do it, I get results that are not consistent with the ones that should occur. The by-hand computed value that the fourth row of the ewma_1 column of the dataframe should be the one that is inside the println statement; 0.15140859291637426 but it gives me the wrong value even if the code suggests it's the same, not only that but the output is inconsistent among different executions of the program. I have not included any random elements to any of the other cells of the notebook or in this one; could this be a vscode problem or am I doing something wrong?
#Inconsistent behaviour of a formula
soft lagoon
ashen nymph
I'm beginning to run the code
but I'm needing to format the code before it can work
Basically whatever the export is doing it's removing all new lines and escaped \
The first cell doesn't format the data for me cos it just says I've already formatted it
In other words it probably went out of bounds
ashen nymph
Ok, the 2nd code block doesnt run cos it assigns to constant_vol which doesn't yet exist
but it prob would work changing .= to =
soft lagoon
Huh
weird, it works in my pc
In what line does the error occur
maybe try to preallocate constant_vol
or just delete the plotting stuff because the problem is the ewma
ashen nymph
Ok I'm starting to guess that it's a list comprehension thing
lazy execution and all maybe
I rewrote it as a for loop and it seems stable
function ewma_094(df)
nrows = nrow(df)
df.ewma_1 = Array{Float64, 1}(undef, nrows)
df.ewma_1[1:99] .= 0.0
df.ewma_1[100] = sqrt( 252 * sum( 0.94^(i-1) * df.Change[(100-i)]^2 for i in 1:99 ) * (1-0.94))
for ridx in 101:nrows
df.ewma_1[ridx] = sqrt( (0.94*(df.Change[ridx]^2) + 0.06*((df.ewma_1[ridx - 1])^2)) * sqrt(252) )
end
return df
end
ewma_094(us_500)```The only real change is actually at the for loop part
I get ```julia
52-element Vector{Float64}:
0.0
0.1452236480399996
0.15262717563008693
0.15140859291637426
0.14951808593099433
0.15862618073364068
0.15846421040896336
0.1555520368563569
0.15206112404377298
0.16526156207840798
0.16139158199426953
0.15816260661792444
0.15533178693338331
our number 0.15140859291637426 is there
Hmm I take it back it's not the list comprehension, but I suspect it's the DataFrames library where assigning to an array utilises multithreading
Then maybe it computes stuff out of order when it has to strictly go in sequence
soft lagoon
kay
soft lagoon
Then maybe it computes stuff out of order when it has to strictly go in sequence
bruh
Okay
It seems to work now
I loaded the data into an array before sending it into the df
Ye it works now
Thanks man
ashen nymph
Sure, thanks