# Bartlett Kernel (Newey West Covariance Matrix)

Referring to Pesaran (2015) the NW covariance matrix is computed according to the following formula:

$$\hat V(\hat\beta)=\frac{1}{T}Q_T^{-1}\hat S_TQ_T^{-1}$$

(Skipping the definition of Q)

$$\hat S_T=\hat\Omega_0+\sum_{j=1}^mw(j,m)(\hat\Omega_j+\hat\Omega_j').$$

(Skipping the definition of $\hat\Omega_j$) Where the Bartlett Kernel is

$$w(j,m)=1-\frac{j}{m+1}.$$

Could someone shed some light on what exactly the Bartlett Kernel "does" for me (and how it "works")? I seem to struggle with understanding the basic theoretical construct that underlies this.

• This may be a starting point: stats.stackexchange.com/questions/153444/… – Christoph Hanck Feb 22 '18 at 7:52
• It is indeed, thank you. I will check out the reference you are giving in the answer and try to work my way through the topic. I think I do understand the basic concept, but the technical details are kind of a black box to me right now. – shenflow Feb 22 '18 at 8:07