¶ 1. Skoglund, J.

The State of Vermont appeals a superior court decision striking down as "arbitrary and capricious" the methodology it used to determine the Town of Killington's statewide school property tax under the Equal Educational Opportunity Act (Act 60), and ordering the Commissioner of Taxes to recalculate the Act 60 tax for Killington. Appellee Town of Killington cross-appeals the court's limited remedy, arguing that the court should have ordered the State to redetermine the 1997 education property tax for all Vermont municipalities. We conclude the trial court erred in invalidating the State procedure, and therefore reverse.

¶ 2. Act 60 created a statewide property tax assessed not upon individual landowners, but against each Vermont municipality, calculated as 1% of the aggregate fair market value of the town's realty. 32 V.S.A. § 5401(6). Because of the disparate appraisal levels and practices in Vermont municipalities, Act 60 requires the Commissioner of Taxes to standardize, or equalize, the aggregate fair market value of each municipality's grand list every year. Id. § 5405(a). The Act provides that the determination of equalized fair market value "shall be based upon such methods, as in the judgment of the commissioner, and in view of the resources available for that purpose, shall be appropriate to support that determination." Id. § 5405(d). A municipality's statewide education property tax liability is the product of the statewide property tax rate multiplied by the municipality's equalized grand list. Id. § 5402(b).

¶ 3. Killington set its 1997 education grand list value at $397,492,895. After applying the State's equalization procedures, the Director of Property Valuation and Review certified Killington's equalized education grand list value as $410,103,965. Id. § 5406(b). Killington petitioned for a redetermination under 32 V.S.A. § 5408, which resulted in a reduction of value to $403,980,885. Killington then appealed the redetermination to the Valuation Appeal Board (VAB). In its decision, the VAB criticized the State's methodology in some categories of property valuation and questioned the credibility of Killington's grand list valuation in general. The VAB ordered Killington to undertake a complete reappraisal of all properties in the town and ordered the State to redetermine Killington's equalized value upon completion of the townwide reappraisal. The VAB ruled that, in the event Killington did not complete a townwide reappraisal within ninety days, Killington's redetermined equalized value would be affirmed.

¶ 4. Killington appealed the VAB's decision to the superior court under § 5408(d), which provides that the court "shall hear the matter de novo." Following an evidentiary hearing, the court ruled in favor of Killington. Notably, the court did not find that the State's estimate of Killington's equalized education grand list value, which differed from Killington's grand list value by only 1.6%, was inaccurate or excessive. Indeed, the trial court reached no conclusion as to Killington's 1997 equalized education grand list value. Rather, the court concluded that the methodologies used by the State in the 1997 equalization study were arbitrary and capricious and that their application constituted an abuse of discretion. Accordingly, the court ordered the State to redetermine Killington's 1997 aggregate fair market value "using statistically appropriate methodologies," but declined Killington's request to order a redetermination of all Vermont municipalities. The court also dismissed five individual intervenors for failure to exhaust administrative remedies. The State's appeal and Killington's and intervenors' separate cross-appeals followed.

I.

¶ 5. It is critical at the outset of our discussion to identify the correct standard of review. Killington notes that the Act authorizes the trial court to hear its appeal from the VAB "de novo," 32 V.S.A. § 5408(d), which signifies that the court shall conduct a new evidentiary hearing and make its own independent findings. State v. Madison, 163 Vt. 360, 369, 658 A.2d 536, 542 (1995). This begs the more fundamental question, however, of the standard of review to be accorded the Commissioner's initial administrative choice of equalization methodology. The Legislature, as noted, specifically provided that the determination of each municipality's equalized education grand list was to be based on "such methods, as in the judgment of the commissioner, and in view of the resources available for that purpose, shall be appropriate to support that determination." 32 V.S.A. § 5405(d). The legislative provision mirrors the substantial deference that courts have traditionally accorded administrative agencies, particularly where, as here, a decision involves highly complicated valuation and equalization methodologies within the agency's area of expertise. As we have observed, "[a]bsent a clear and convincing showing to the contrary, decisions made within the expertise of . . . agencies are presumed correct, valid and reasonable." In re Johnston, 145 Vt. 318, 322, 488 A.2d 750, 752 (1985) (citing Vermont Dep't of Taxes v. Tri-State Indus. Laundries, Inc., 138 Vt. 292, 294, 415 A.2d 216, 218 (1980)).

¶ 6. Thus, to successfully challenge the Commissioner's equalization methodology, it was Killington's burden here to demonstrate that the State's approach was wholly irrational and unreasonable in relation to its intended purpose. See, e.g., Breault v. Town of Jericho, 155 Vt. 565, 569, 586 A.2d 1153, 1156 (1991) ("If the [Board of Appraiser's] decision is within the range of rationality, it must be affirmed."); Sondergeld v. Town of Hubbardton, 150 Vt. 565, 568, 556 A.2d 64, 66 (1988) ("The burden of proof is not met by simply impugning the Board's methods or questioning its understanding of assessment theory or technique. To prevail a taxpayer must show an arbitrary or unlawful valuation."). Requiring such a relatively high threshold of proof ensures the necessary flexibility and range of judgment to which the Commissioner is entitled under the Act, while protecting against the purely aberrational or capricious result.

¶ 7. Turning to the record evidence and the trial court's decision, several initial observations are in order. There is no doubt that the depth and breadth of the statistical evidence adduced by the parties was impressive. Nevertheless, certain fundamental facts appear to have been lost in the blizzard of statistical data and expert opinion. First, it is clear that the State calculated Killington's equalized education grand list value in accordance with its established procedures. The State analyzed 187 bona fide sales within the Town of Killington during the previous two years, which represented over six percent of all real properties in the Town, well in excess of the four percent minimum required by the State. Using these transactions, the State calculated the assessment level, or ratio, of sales price to listed value for each category of property, derived an equalized value, and added those values to yield an aggregate fair market value within 1.6% of Killington's grand list value. Killington adduced no credible evidence to suggest that the fair market value of its equalized education grand list differed markedly from its own assessment figure or that of the State.

¶ 8. Second, the court made no findings, and indeed took virtually no account of the State's evidence relating to the reasonableness of its equalization methodology in light of the resources available, as required by Act 60. See 32 V.S.A. § 5405(d) (determination of equalized fair market value "shall be based upon such methods, as in the judgment of the commissioner, and in view of the resources available for that purpose, shall be appropriate to support that determination"). (Emphasis added.) Thus, while critical of the limited sales data utilized by the State and the fact that the State failed to supplement the data with independent appraisals, the court largely ignored the testimony of state officials that appraisals were beyond the limited means available to the Division of Property Valuation and Review, which had devoted the bulk of its first-year budget to improving local listing practices by purchasing computers for towns and holding training sessions for local listers. This was based upon the Director's considered judgment that such efforts would yield more accurate equalization results in the longterm, and nothing adduced by Killington shows that decision to have been wholly unreasonable.

¶ 9. Third, it is uncontested that the State's estimate of Killington's equalized education grand list had an overall confidence interval of plus/minus 4.6%, which is generally acknowledged to be a reliable result. Fourth, the State presented empirical evidence based on a wide-ranging study of 134 towns that had reappraised between 1991 and 1996, showing a close correlation between the reappraised values and the State's equalized education grand list values and indicating that the State's equalization methodology is an accurate predictor of fair market value. Finally, the State's expert, Robert J. Gloudemans, an acknowledged leader in the field of appraisal procedures and the author of several standard reference works, testified that the State's equalization methodology was reasonably consistent with accepted industry standards and that Killington's equalized education grand list value was a reasonably reliable estimate.

¶ 10. The record thus contains ample evidence to support a conclusion that the equalization methodology which the Commissioner employed to calculate Killington's aggregate fair market value was rational, and yielded a reasonably reliable result. This was sufficient to affirm the decision of the VAB. Breault, 155 Vt. at 569, 586 A.2d at 1156. Although, as discussed more fully below, Killington adduced conflicting evidence critical of the State's procedures, a proper deference to the Commissioner's reasoned judgment requires that we uphold her chosen methodology if credible evidence shows that it was rational and reliable. Lake Morey Inn Golf Resort, Ltd. P'ship v. Town of Fairlee, 167 Vt. 245, 248-49, 704 A.2d 785, 787 (1997). It is not our function, in cases such as these, to choose between competing statistical models.

II.

¶ 11. The State's evidence to the contrary notwithstanding, Killington claimed, and the trial court found, that the equalization methodology employed by the Commissioner was fatally deficient in several respects. To analyze the claim, a brief summary of that methodology is in order. In 1997, the State equalized each municipality's grand list by collecting data on real estate transactions that occurred in the municipality from April 1, 1995 through March 31, 1997. The State then screened the sales, with the assistance of the municipality's listers, to derive a list of bona fide, arms-length sales and calculated a listed-value-to-sale-price ratio for each sold property. The State then placed each sale into one of fifteen property categories and calculated an applied ratio for each category, used the ratio to derive an equalized value for all properties in the category, and added the values to obtain the aggregate fair market value of the municipality's grand list.

¶ 12. At trial, Killington challenged several aspects of the State's methodology. It claimed, and the court found, that the requirement of at least three sales within a given category of property was statistically inadequate, and yielded unacceptably large confidence intervals (which the court defined as exceeding five percent). The court cited an independent study of the State's equalization procedures (the so-called Almy Report) which the Tax Department had commissioned on the recommendation of the legislative Commission on Property Tax Appraisals. Other evidence, however, indicated that three comparable sales is sufficient to support a reasonable estimation of a given category's assessment level, particularly where, as in Vermont, reliance on a large number (fifteen) of property categories results in a higher degree of homogeneity within categories. The trial court was also critical of the relatively high confidence levels within several categories of property, concluding that they rendered Killington's aggregate fair market value unreliable. The court inexplicably discounted, however, the undisputed evidence that the category with the vast majority of sales (152 out of 187) — vacation homes on six acres or less — had an extremely reliable confidence interval of 2.1%, and State experts and officials, including Mr. Gloudemans, testified that what counted was the "bottom-line number," not confidence intervals for smaller property categories which did not significantly affect the overall result.

¶ 13. Killington also challenged several aspects of the State's methodology relating to the placement of property within categories. It claimed, and the court found, that inclusion of the Killington Ski Area in the commercial category was irrational because the category consisted predominantly of small businesses. Yet the court could equally have found that Killington's small restaurants, inns, and retail shops were related to the ski industry and subject to similar market trends. The court also criticized the State's inclusion of different types of undeveloped Killington property within the "miscellaneous" category, citing testimony that shoreland and wasteland will fare differently in the marketplace, yet the court made no finding concerning the impact of this limited category on Killington's overall aggregate fair market value. Additionally, the court criticized the State's practice of combining sales from separate categories when individually they lacked sufficient sales to derive an equalization ratio, a practice characterized by the Almy Report as "dubious." Yet it is not uncommon to use sales from different classes of property to determine an equalization ratio where a statistically representative sample is lacking and the categories share market characteristics. See Philbin v. Town of St. George, 156 Vt. 640, 641, 588 A.2d 1060, 1061 (1991) (mem.) (where Board concludes that it lacks evidence of statistically representative sample to determine equalization ratio, it may use evidence of other classes). Furthermore, there was no showing that the single instance cited by the court — combining one sale from the woodland category with twelve sales in the miscellaneous category — had any significant impact on the Town's aggregate fair market value.

¶ 14. The trial court was also critical of the State's reliance on sales from the two years prior to the appraisal date, noting that studies that had utilized sales during the one year before and one year after the appraisal date were more accurate. Again, however, there was no clear evidence that the failure to adjust for sales more than one year old had any significant effect on Killington's aggregate fair market value. The court further disparaged the State's use of the so-called O'Brien Table to determine an applied ratio when extremes among sales within a category made use of an aggregate ratio unreliable. The court cited a recommendation in the Almy Report that the O'Brien Table be replaced by some other more recognized measure of central tendency. Yet here again there was no substantial evidence that the State's reliance on the O'Brien Table — which according to the Commissioner had effectively compensated for sales data skewed toward high- or low-priced properties — had any significant effect on the State's estimate of Killington's aggregate fair market value.

¶ 15. The trial court also found that the State's method for determining the equalized value of utility properties, which involved a procedure for indexing earlier determined fair market values upward when the property represented less than five percent of total property value within a municipality, was "substantially flawed" because it did not employ one of three recognized appraisal methods (cost, sales comparison, and income) and assumed rising utility values. The court's finding accurately reflects statements in the Almy Report and testimony by Killington's expert. Yet the court made no finding as to the actual predictive value of the State's procedure, ignoring evidence that Killington's reassessed 1998 utility value increased to within 95% of the State's estimate; no finding that the equalized value of utilities in Killington or elsewhere differed significantly from the State's estimate; and no finding that the utility value significantly affected Killington's aggregate fair market value.

¶ 16. Finally, the court found that the State's equalization methodology failed to assure a reasonably proportioned statewide property tax because a significant number of municipalities had high coefficients of dispersion (COD). See Knollwood Bldg. Condos. v. Town of Rutland, 166 Vt. 529, 540, 699 A.2d 31, 39 (1997) ("The extent to which a grand list achieves equity among taxpayers is commonly measured by the `coefficient of dispersion,'" which measures the percentage of deviation of individual assessment or equalization ratios from the general average ratio for the area); 32 V.S.A. § 5401(1) (defining process for determining municipality's COD). The evidence showed, however, that Killington had a COD of slightly over eleven percent, which is generally acknowledged to be reliable, and that the majority of Vermont municipalities had statutorily acceptable CODs of twenty percent or under. See 32 V.S.A. § 4041a(b) (municipality with certified COD greater than twenty must reappraise its education grand list properties). Accordingly, the evidence did not support the court's finding that significant numbers of municipalities with high CODs invalidated the State's equalization study. Furthermore, while the court relied on Mr. Gloudemans's testimony that a high COD necessarily affects the State's ability to equalize aggregate fair market value reliably, it overlooked his additional testimony that the way to improvement lay not in the State's equalization methodology, but in the towns' efforts to reappraise.

¶ 17. The record evidence considered in its entirety does not, in sum, support the court's findings invalidating discrete portions of the State's equalization methodology, and the findings — considered individually or in combination — do not support the court's overall conclusion that the State procedure was arbitrary and capricious, or yielded unreliable results. On the contrary, the record reveals that the State adduced ample credible evidence demonstrating that its methods — while limited by the resources available and improvable in certain areas — comported with industry standards and yielded a reasonably reliable estimate of aggregate fair market value. Accordingly, the trial court judgment invalidating the State's Act 60 equalization methodology was erroneous and must be reversed. In light of this conclusion, it is unnecessary to address the State's alternative assertion that Killington lacked standing to challenge the State's methodology, or to consider the cross-appeals of Killington and the individual intervenors.

Reversed.