Here's a copy of my output:

Algorithm converged.

GARCH Estimates

SSE 4.00584241 Observations 74

MSE 0.05413 Uncond Var 0.05878857

Log Likelihood 8.70145121 Total R-Square 0.7052

SBC 42.8540089 AIC 10.5970976

MAE 0.18238075 AICC 17.7157416

MAPE 8.30535866 Normality Test 0.5139

Pr > ChiSq 0.7734

Standard Approx Variable

Variable DF Estimate Error t Value Pr > |t| Label

Intercept 1 1.5872 0.1002 15.84 <.0001

pmmood 1 0.0165 0.003971 4.16 <.0001 pmmood

rundummy 1 0.2330 0.0551 4.23 <.0001

sndyfndy 1 -0.000319 0.0000795 -4.02 <.0001

lnghtdv 1 0.1442 0.0586 2.46 0.0138 lnghtdv

bshawhrs 1 0.1044 0.0301 3.47 0.0005 bshawhrs

kellihrs 1 0.0249 0.007898 3.16 0.0016 kellihrs

bakehrs 1 0.1295 0.0390 3.32 0.0009 bakehrs

musichrs 1 0.0275 0.008310 3.31 0.0009 musichrs

pytave 1 0.2596 0.0597 4.35 <.0001 pytave

AR1 1 0.7024 0.1527 4.60 <.0001

AR2 1 0.3699 0.1182 3.13 0.0018

ARCH0 1 0.0175 0.0199 0.88 0.3796

ARCH1 1 0.5396 0.3230 1.67 0.0948

GARCH1 1 0.1821 0.4284 0.43 0.6707

Thanks for any help. I know the sample size is way too small for a GARCH model, but this is a preliminary run of data...there'll be a few thousand observations once the collection is done at the end of the year.

I found an HC3 macro to use on the small sample to deal with the heteroskedasticity, but a) I can't figure out how to also deal with autocorrelation when using the macro; and b) I want to account for volatility, not necessarily have homoskedastic errors. If anyone thinks there is a better way to do this (dealing with the heterosk., feel free to let me know.

As it stands, I can't find any reliable, not absurdly technical explanation of dealing with the coefficients that the GARCH process adds to the output.

Thanks for the help