"But how can I find out the overall queue size that WRED is being used against?"
Cannot say for sure, as I don't have a C9200 to "play" with, and how and what stats Cisco devices show aren't always exactly the same, for same command, between devices, but, have you tried a show policy interface?
"Also, is there a way to determine if the threshold values I chose are working well or not?"
Depends on how you define "working well".
"Any recommendations on initial threshold values?"
Depends on your traffic and your QoS goals.
(Sorry if my last two answers seem to be ducking your questions, but using RED optimally requires a deep understanding of your traffic's attributes and what you want to accomplish using RED.)
When I wrote in prior post "I generally advise, unless you're a QoS expert, don't use WRED.", I'm not joking.
The way RED (WRED) is simply explained, it seems such a great approach and easy to use to obtain wonderful results. It can be very effective, but it's (surprisingly) difficult to get it to work ideally (even if you're a QoS expert).
Without going into all the issues I believe you can run into using RED, (which many later variants "try" to fix or mitigate), "typical" network applications, and their "expectations" of network service, might just differ a bit since 1993, which is when Dr. Floyd proposed RED. (Also since '93, network hardware is, generally, more "powerful", and "smarter" approaches can be used rather than RED.)
With LAN devices, e.g. LAN switches, often many congestion issues, beyond dealing with really sensitive traffic, e.g. VoIP, are dealt with by using more bandwidth.
Where bandwidth is more more problematic to obtain (or afford), routers, with their more full featured QoS, can do QoS "better" than a LAN switch.
