SPML: Error in lag.listw(listw, u, zero.policy = zero.policy) : object lengths differ

Question

I have panel data and need to apply spml but I keep getting the same mistake and despite trying I cannot manage to overcome it.

Basically, I have a balanced panel of 36 units and 261 days of observations. I created a data.frame (named aaa) with 6 columns

Bank, Date, CDS, X1, X2, X3

and 9396 rows (9397 if considering the column labels).

Here: * Units and Date are my indexes * CDS is the dependent variable * X1, X2 and X3 are the independent variables.

• Bank is a Factor w/ 36 levels,
• Date is an Integer vector int [1:9396],
• CDS, X1, X2 and X3 are all numeric vectors num [1:9396].

I also created a listw object (named W2) of class “matrix” and structure num [1:9396, 1:9396].

I obtained my listw object by applying Kronecker to an initial matrix W with the following code:

(Kronecker (diag(261), W)

I then expressed the “Formula” in this way:

fm <- (Cds) ~ (X1) + (X2) + (X3)

And, subsequently applied splm:

spmltest<-spml(fm, data = aaa, index = c("Bank","Date"), listw = w2)

Here’s the error I get:

Error in lag.listw(listw, u, zero.policy = zero.policy) : object lengths differ.

I tried to check but seems to me that objects do not actually have different lengths.

Any idea where I am mistaken? Any help or suggestion would be much appreciated. Thank you.

(Ps: I know I should specify the model (if fixed or random) as well, but I don’t think this would solve this specific issue. If I am wrong, please let me know)

(Pss: I also found that someone already asked about the same issue and solved this way: “The authors of the package "splm" pointed out that there was a mistake in the way the panel ids were set up. The spatial units must be slower (1,1,1,.....; 2,2,2...;....;77,77,77,....) and the time id must be faster (1,2,3,....,31;....; 1,2,3...31) of the two. After setting up the data accordingly, I could get the results”. I just cannot understand what it is meant with slower and faster. If that means that data should be sorted by spatial units rather than by time, my data are already sorted this way).

Answer 1

My experience is that I got the same error when I was using panel containing NA values. When I remove the data with NA, the code was working properly. Good luck!

Answer 2

I had the same problem in my model.

I solved (special thank to Millo Giovanni) by setting the first two columns of my database by "id" and "time". You also do not need to proceed the kronecker product, R aumatically expands the spatial matrix W.

And do not forget to identify the data in the same order, "id" and "time" by index = c("id", "time").

There is my code:

SEM <- spml(formula = gdp ~ oivrev, data = data, index = c("id", "year"), listw = W, model = "within", lag = F, spatial.error = "b")

I hope it helps you.

Answer 3

you need to set index=Null since you do Kronecker product