These appeals arise out of District Judge Harold H. Greene’s rulings against the Department of the Navy (“Navy”), an operational unit within the Navy, and certain named officials in four consolidated Title VII1 sex discrimination cases on the ques*360tions of liability,2 relief,3 and reconsideration.4 We are asked to review the District Court’s conclusion that the appellants discriminated against a class of female professional technical employees in initial grade placements and promotions. We are also requested to review the District Court’s ruling that, once an individual class member shows in a remedial hearing that she was a potential victim of the proven discrimination, the burden shifts to the employer to demonstrate by clear and convincing evidence that its employment decisions with regard to that individual were based on legitimate factors unrelated to the policy of discrimination that has already been proved. Finally, the appellants challenge the District Court’s findings of discrimination against two individual employees and the relief ordered for those individuals.
In pursuing these appeals, the appellants have raised an issue concerning the sufficiency of the prima facie case presented by the class plaintiffs. On the record at hand, we believe that the appellants’ objections on this score arise from too stringent a view of the requisites of a prima facie case. “The burden of establishing a prima facie case of disparate treatment,” the Supreme Court has emphasized, “is not onerous.” Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 253, 101 S.Ct. 1089, 1093, 67 L.Ed.2d 207 (1981). On the basis of the evidence of discrimination introduced at trial, Judge Greene properly held that the class plaintiffs had satisfied this burden; the Navy’s objections to the plaintiffs’ statistical presentation were speculative and unqualified, and the court’s decision to disregard those challenges was neither clearly erroneous nor contrary to established legal principles.5 We hold, however, that evidence introduced at trial effectively rebutted the inference that the defendants had discriminated in initial grade placements, and we reverse that aspect of the District Court’s judgment of class-wide liability.
The remaining issues raised in these appeals are less difficult to resolve. As the appellants appeared to concede in the oral argument before this court, the District Court’s application of the clear and convincing evidence standard in individual relief hearings is consistent with the law of this circuit. The District Court’s findings of discrimination against two individual employees, moreover, rest largely on its evaluation of the design, motive, and intent with which the appellants acted; while “it would be our duty to correct clear error,” United States v. Yellow Cab Co., 338 U.S. 338, 342, 70 S.Ct. 177, 179, 94 L.Ed. 150 (1949), the trial court’s choice between two permissible views of the weight and direction of the evidence was not clearly erroneous. Nevertheless, we are unable to determine whether the District Court acted within its authority in fashioning relief for these individuals, and we remand several elements of the relief award for further consideration.
I. Background
A. Factual Background
The appellants in these cases are officials of the Navy and an agency of the Navy formerly referred to as the Naval Command Support Activity and now known as the Navy Regional Data Automation Center (“NAVCOSSACT/NARDAC”). NAVCOSSACT was originally established as the center of Navy computer operations for national defense purposes; in 1977, NAVCOS-SACT was dissolved and its personnel and *361resources were combined with those of several other agencies to form NARDAC, which develops computer systems and documents for computer systems, delivers the systems and documents to user organizations, and trains personnel in their operation. The appellees are employees or former employees of NAVCOSSACT/NARDAC, some of whom are members of the class of “all female professional technical employees employed by ... [NAVCOS-SACT] or ... [NARDAC] at any time between June 6, 1972, and June 4, 1979.”6
The genesis of this protracted litigation, the annals of which fill many volumes and the resolution of which already has consumed a large chunk of judicial resources, can be traced to June 21,1972. It was then that Yvonne Trout, a NAVCOSSACT computer systems analyst, filed an informal charge of sex discrimination on behalf of herself and a class of female NAVCOS-SACT employees. This charge, the relevant portion of which alleged “[t]he existence of a continuing and pervasive pattern of discrimination against women which results in their being deprived of consideration for and promotion to upper grade levels and management positions,” II Jt.App. 632-33, was made part of a formal complaint filed by Trout on September 20, 1972. Shortly thereafter, on January 10, 1973, Trout filed a civil action in the District Court alleging, in part, “a continuing and pervasive pattern of discrimination with respect to all professional technical female applicants and employees.” Complaint ¶ 63, R.l.
For purposes of this appeal, the extensive procedural history of Trout’s civil action may be summarized quickly. In April 1973, Clara Perlingiero, another NAVCOSSACT computer systems analyst who had filed informal charges of sex discrimination in September 1971 and August 1972 and a formal administrative complaint in September 1972, was allowed to join as a party plaintiff. Immediately thereafter, the District Court conditionally certified a class consisting of “all past, present and future female professional technical employees of ... [NAVCOSSACT].” Trout v. Warner, Civ. No. 73-55 (D.D.C. Apr. 6, 1973) (order authorizing class action), R.16. In 1976, two additional cases involving sex discrimination were filed by NAVCOSSACT employees and consolidated with Trout’s case; one of these suits alleged discrimination against Marie Bach, a security manager and nonmember of the conditionally certified class. An independent action filed by Trout was also consolidated with her original case in 1978. The conditionally certified class was redefined in June 1979, see text at note 6 supra, and a ten-day trial of the consolidated actions was conducted in June 1980.
B. The District Court’s Decisions
1. Class Claims
As the District Court recognized, “[t]he class action aspects of this lawsuit involve an alleged pattern and practice of sex discrimination in [NAVCOSSACT/NARDAC’s] hiring, performance evaluation, job assignment, promotion, and award procedures.” Trout v. Hidalgo, 517 F.Supp. 873, 877 (D.D.C.1981). Hiring and initial job placement procedures, however, were the subject of only a small fraction of the trial testimony and exhibits. The meager evidence adduced regarding the determination of initial grade levels for new NAVCOSSACT/NARDAC employees uniformly supported the defendants’ contention that they were not responsible for initial placement decisions. The plaintiffs, in fact, made no effort whatsoever to rebut the defendants’ evidence attributing sole responsibility for initial placement decisions to the Civil Service Commission (“CSC”) and the Office of Personnel Management (“OPM”). Notwithstanding this deficiency in the plaintiffs’ proof with respect to initial placements, the District Court held, on the basis of extensive statistical evidence and the testimony of a number of present and former NAVCOSSACT/NARDAC employees, that the class plaintiffs had established a prima facie case of sex discrimina*362tion in both initial placements and promotions, and that the defendants had failed to rebut that showing.
These conclusions rested primarily, but not exclusively,7 on the District Court’s assessment of the probative value of the competing statistical studies presented by the parties. The plaintiff’s expert, whose techniques and findings were ultimately accepted by the court, began his analysis with the undisputed salary differentials between men and women at NAVCOSSACT/NARDAC, reflecting a relative concentration of women in lower-level positions. His study was designed to determine whether these differentials could be explained by differences between men and women in educational attainment and experience. Using data from NAV COSSACT/NARDAC personnel records, he specified a model in which the dependent variable was salary and the independent variables were (1) level of education, (2) years of NAVCOSSACT/NARDAC experience, (3) years of other governmental experience, (4) years of potential nongovernmental experience between the date of receipt of last educational degree and date of entry into the federal service, and (5) sex. Factors (1) through (4) proved insufficient to explain the salary differentials. Even when differences in education and experience were taken into account, female employees continued to receive substantially lower salaries than men. Indeed, “[t]hese differences in salaries were statistically significant at the .01 level, that is, they would be expected to arise by chance less than one percent of the time if there were no discrimination based on sex.” Trout, 517 F.Supp. at 879 n. 14. An additional regression confined to employees hired after 1972, the effective date of Title VII in the federal service, yielded results consistent with the basic model.
The defendants responded to this statistical showing of discrimination in two ways. First, they argued that the plaintiffs’ statistics were insufficiently refined to support a prima facie case because (1) the regressions omitted critical variables concerning the minimum objective qualifications required for promotions within NAVCOS-SACT/NARDAC, (2) the regression results were biased by the expert’s failure to eliminate the effects of nonactionable pre-1972 discrimination, and (3) the inclusion of the effects of discrimination by agencies other than NAV COSSACT/NARDAC skewed the regression results. The view of the requisites of a prima facie case embodied in these objections was rejected by the trial judge, who apparently believed that while otherwise valid studies incorporating these factors could be used to rebut a prima facie case, the plaintiffs’ statistical evidence was sufficiently accurate to give rise to an inference of actionable discrimination. Such studies did constitute the second prong of the defendants’ response to the plaintiffs’ case, but the District Court concluded that the statistical analyses proffered by the defendants suffered from a number of defects that rendered them less persuasive than the plaintiffs’ exhibits.
Following the trial of the plaintiffs’ liability claims, and after holding the defendants liable for discrimination against the class of female professional technical employees at NAVCOSSACT/NARDAC, the District Court directed the parties to submit memoranda on potential methods of determining individual employees’ entitlement to relief. Taking liberties with this instruction, the defendants presented a number of new statistical studies on the question of class liability and substantial documentary evidence pertaining to their responsibility for initial grade placement decisions. These data later formed the basis for a motion for reconsideration, which the trial court characterized as “more akin to a demand for a new trial than for reconsidera*363tion.”8 The basic thrust of this motion was that the defendants, in preparing for the relief proceedings, had collected new evidence that required the court to enter judgment for them on the question of liability. In the court’s view, however, this motion was untimely and provided no. basis for either reconsideration or a new trial. The court reasoned that the defendants had had ample time to prepare their case, the evidence was not of a sort that would have been unavailable to a duly diligent defendant at the time of the trial, the new studies were cumulative and unlikely to affect the result, and the resolution of the entire matter would not be advanced by another trial on the question of liability.
If the District Court’s rulings on class liability are affirmed, this lawsuit must enter a second phase, in which class members’ entitlement to relief will be determined on an individualized basis. See, e.g., International Brotherhood of Teamsters v. United States, 431 U.S. 324, 360-62, 97 S.Ct. 1843, 1867-68, 52 L.Ed.2d 396 (1977). On this point, the District Court held that, after an employee demonstrates in remedial hearings that she was a potential victim of the proven discrimination, “the burden shifts to the employer to demonstrate by clear and convincing evidence that its employment decisions with regard to that individual were based on legitimate factors unrelated to the policy of discrimination which has already been proved.”9 This placement of the burden of persuasion provides a second ground for challenging the District Court’s resolution of the class action aspects of this suit.
2. Individual Claims
In addition to resolving the issue of class liability, the District Court considered five individuals’ allegations of sex discrimination and ruled in favor of two of the employees. It found, first, that Marie Bach, a nonmember of the class, had been removed from her position as a GS-13 security manager in 1976 and 1979, replaced by younger males with grades of GS-14 and GS-15, and subjected to other discriminatory acts, including the failure to classify her position at a grade comparable to that of other Navy security specialists with similar functions. Among the relief awarded Bach were two retroactive promotions and commensurate backpay. The court also found that Clara Perlingiero, a class member, had been discriminatorily denied promotions in 1971 and 1979 and subjected to other discriminatory acts, including the failure to classify her position at a grade comparable to similarly situated males. Among the relief awarded Perlingiero were retroactive promotions to GS-13 as of September 1971 and to GS-14 as of September 1979 and commensurate backpay. Both the findings of liability and the awards of relief have been challenged in these appeals.
II. Discussion
A. The Scope of Review
In reviewing the challenged decisions, we are guided by the principle that “factfinding is the basic responsibility of district courts, rather than appellate courts,” DeMarco v. United States, 415 U.S. 449, 450 n. *, 94 S.Ct. 1185, 1186, n. *, 39 L.Ed.2d 501 (1974) (per curiam), and the rule that a district court’s factual findings may not be set aside unless clearly erroneous, Pullman-Standard v. Swint, 456 U.S. 273, 287, 102 S.Ct. 1781, 1789, 72 L.Ed.2d 66 (1982); Fed.R.Civ.P. 52(a). This rule, of course, does not apply to conclusions of law, Pullman-Standard, 102 S.Ct. at 1789, or to factual findings that result from the application of incorrect legal principles, Inwood Laboratories, Inc. v. Ives Laboratories, Inc., 456 U.S. 844, 855, 102 S.Ct. 2182, 2189 n. 15, 72 L.Ed.2d 606 (1982). But most findings of fact, including findings of discriminatory intent, will be left undisturbed unless “the reviewing court on the entire evidence is *364left with the definite and firm conviction that a mistake has been committed.” United States v. United States Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 542, 92 L.Ed. 746 (1948).
Although many of the issues raised in these appeals involve purely factual conclusions of the District Court, several present legal questions or mixed questions of law and fact. The trial court’s findings of discrimination against Bach and Perlingiero and its assessment of the relative probative force of the parties’ statistical evidence, for example, are undoubtedly factual. See Pullman-Standard, 102 S.Ct. at 1790-91; Medina v. Reinhardt, 686 F.2d 997, 1007 (D.C.Cir.1982). But the appellants’ challenges to the sufficiency of the prima facie case of discrimination against the class of female professional technical employees at NAV COSSACT/NARDAC, grounded as they are in the alleged omission from the statistical showing of factors reflecting important legal principles, raise legal issues that trigger a more careful examination by this court.
B. The Prima Facie Case
To establish a prima facie case of sex discrimination in initial grade placements or promotions, plaintiffs are required to show only that it is more likely than not that their employers failed properly to place or promote women because of discriminatory intent. See Furnco Construction Corp. v. Waters, 438 U.S. 567, 576, 98 S.Ct. 2943, 2949, 57 L.Ed.2d 957 (1978); O’Brien v. Sky Chefs, Inc., 670 F.2d 864, 866 (9th Cir.1982). This burden is not a heavy one, Texas Department of Community Affairs v. Burdine, 450 U.S. at 253, 101 S.Ct. at 1093, and the required showing may be made by statistics alone, Hazelwood School District v. United States, 433 U.S. 299, 307-08, 97 S.Ct. 2736, 2741-42, 53 L.Ed.2d 768 (1977), or by a “cumulation of evidence, including statistics, patterns, practices, general policies, or specific instances of discrimination,” EEOC v. American National Bank, 652 F.2d 1176, 1188 (4th Cir.1981), cert. denied, - U.S. -, 103 S.Ct. 235, 74 L.Ed.2d 186 (1982).
If the plaintiffs have used statistics to make out a prima facie case, the defendants may rebut the inference of discrimination by presenting a legitimate, nondiscriminatory explanation for the observed statistical disparity or by raising a genuine issue of material fact concerning the accuracy of the picture painted by the plaintiffs’ statistics. See International Brotherhood of Teamsters v. United States, 431 U.S. at 360, 97 S.Ct. at 1867; Croker v. Boeing Co. (Vertol Division), 662 F.2d 975, 991 (3d Cir. 1981) (en banc). To prevent the latter course from becoming an incomprehensible battle of the experts, several general principles should guide a court’s assessment of the sufficiency of the plaintiffs’ prima facie case. The most important of these principles is that, while plaintiffs must demonstrate to the court’s satisfaction that their statistical comparisons are meaningful, they need not present a perfect statistical analysis at the prima facie case stage. See Dothard v. Rawlinson, 433 U.S. 321, 331, 97 S.Ct. 2720, 2727, 53 L.Ed.2d 786 (1977); Medina v. Reinhardt, 686 F.2d at 1008-09; Vuyanich v. Republic National Bank, 505 F.Supp. 224, 306-07, 354-57 (N.D.Tex.1980).
The appropriate degree of refinement of the plaintiffs' statistical analysis, moreover, may depend on the quality and control of the available data. See Vuyanich, 505 F.Supp. at 356. If the plaintiffs account for the effects of extraneous variables to the extent reasonably permitted by the available data and the evidence presented strongly supports an inference of discriminatory treatment, the District Court properly may conclude that the plaintiffs have made out a prima facie case. As noted in Detroit Police Officers’ Association v. Young, 608 F.2d 671 (6th Cir.1979), cert. denied, 452 U.S. 938, 101 S.Ct. 3079, 69 L.Ed.2d 951 (1981), a court should not “require proof to a mathematical certainty, ... [because] there is no such requirement [under Title VII]. Deficiencies in the data base ‘may, of course, detract from the value of [statistical] evidence,’ but ordinarily would not obliterate its evidentiary value.” Id. at 687 (citations omitted) (quoting *365Teamsters, 431 U.S. at 340 n. 20, 97 S.Ct. at 1857 n. 20). This is particularly true where, as here, the defendants controlled the only sources of data on which statistical analyses could be based. As the District Court observed, “plaintiffs cannot legitimately be faulted for gaps in their statistical analysis when the information necessary to close those gaps was possessed only by defendants and was not furnished either to plaintiffs or to the Court.” Trout, 517 F.Supp. at 883 (footnote omitted).
Furthermore, the most effective way to rebut a statistically based prima facie case is to present more accurate statistics. As the Supreme Court has made clear, where “the employer discerns fallacies or deficiencies in the data offered by the plaintiff, he is free to adduce countervailing evidence of his own.” Dothard, 433 U.S. at 331, 97 S.Ct. at 2727. But unquantified, speculative, and theoretical objections to the proffered statistics are properly given little weight by the trial court:
When a plaintiff submits accurate statistical data, and a defendant alleges that relevant variables are excluded, defendant may not rely on hypothesis to lessen the probative value of plaintiff’s statistical proof. Rather, defendant ... must either rework plaintiff’s statistics incorporating the omitted factors or present other proof undermining plaintiff’s claims.
Segar v. Civiletti, 508 F.Supp. 690, 712 (D.D.C.1981). Although NAVCOSSACT/NARDAC did present some studies of its own — ultimately rejected by the District Court — tending to undermine the plaintiffs’ claims, it did not, for the most part, attempt to correct the alleged deficiencies in the plaintiffs’ statistical presentation in its trial testimony or exhibits. See IV Jt.App. 1921-23. For this and other reasons — set forth below — these alleged shortcomings neither prevented the plaintiffs from establishing a prima facie case nor constituted a persuasive rebuttal to an otherwise adequate initial showing.
1. Minimum Objective Qualifications
The appellants’ principal objection to the statistical analyses on which the District Court relied appears to be that the statistics ignored the minimum objective qualifications necessary for placement and promotion in the professional technical job series at issue. In Davis v. Califano, 613 F.2d 957 (D.C.Cir.1979), this court suggested that “the minimum objective qualifications necessary for one to be eligible for promotion must be considered in the statistical data presented initially by a plaintiff.” Id. at 964 (emphasis in original). The Davis court also speculated that “if a particular number of years of work experience were established as a minimum job criterion, then that would need to be reflected in the proffered statistics.” Id. At least part of the Davis court’s rumination became the law of this circuit in Valentino v. United States Postal Service, 674 F.2d 56 (D.C.Cir. 1982), in which we affirmed a ruling that members of a class consisting of a wide variety of occupational categories had failed to make out a prima facie case of sex discrimination. “ ‘When special qualifications are required to fill particular jobs,’ ” the court declared, “proof that does not center on those ‘who possess the necessary qualifications’ falls short.” Id. at 67-68 (quoting Hazelwood School District v. United States, 433 U.S. at 308 n. 13, 97 S.Ct. at 2742 n. 13).10
Relying on these cases, the appellants claim that placements and promotions within the NAV COSSACT/NARDAC work force rest on minimum objective qualifications established by the OPM. These qualifications, they argue, can be summarized in terms of years of computer-related experience; as a result, any analysis of initial placements and promotions above the GS-5 *366level must include the amount of computer-related experience possessed by the employees. Because the appellants contend that the plaintiffs’ statistics did not account for such experience, they conclude that the regressions could not establish a prima facie case. We reject this line of argument for two principal and related reasons: first, we find that the plaintiffs’ statistical analyses did account for minimum objective qualifications; and second, contrary to the appellants’ assertions, we find the conclusions reached by the District Court in this case to be consistent with the principles enunciated in Valentino.
Not only did the Valentino court implicitly approve the results reached by the District Court in this case, see id. at 68, 70, but the reasoning underlying its decision is wholly inapplicable in a case, like this one, where all members of the class are professional technical employees with generally similar job skills. The regression analyses under consideration in Valentino, like the studies presented in this case, attempted to estimate the effect of sex, years of education, and length of governmental service on employee’s salaries. But that case, unlike this one, involved a class composed of individuals in numerous occupational categories with readily distinguishable qualifications whose experts had “utterly failed to control for type of education and job classification.” Id. at 70. The “grossly imprecise” comparisons that resulted, Medina v. Reinhardt, 686 F.2d at 1010, simply could not provide a reasonable basis for inferring disparate treatment. It would, it seems clear, truly have been “irrational to assume ‘equal qualifications’ to fill engineering or secretarial vacancies among persons educated the same number of years and employed by the government for the same length of time.” Valentino, 674 F.2d at 71 (footnotes omitted).
So understood, the result in Valentino turns on the inadequacy of the proxies— years of education and government experience — for the myriad minimum objective qualifications of the diverse positions held by class members. No such problem is present here. The class certified by the District Court consists solely of employees performing computer-related tasks, and the only minimum objective qualification identified by the appellants is one-to-three years of computer-related experience. Far from being “grossly imprecise,” the plaintiffs’ proxy for this qualification — years of experience with NAVCOSSACT/NARDAC— provides a perfectly adequate basis for inferring disparate treatment. As the appellants observe, this proxy arguably does equate time spent in a clerical position with time spent in the computer field and could thus penalize them for moving female employees from lower-level, nonprofessional jobs into technical professional computer positions with higher potential salaries. Brief for Appellants at 44 & n. 21. But the record does not suggest that such movement occurred with any regularity, see IV Jt.App. 1860-61, much less with sufficient frequency to cast doubt upon the inference of discrimination arising from the plaintiffs’ statistical showing.11
2. Relevant Time Frame
The appellants also argue that the plaintiffs’ statistics failed to make out a *367prima facie case because they ignored the time limitations embodied in Title VII. Absent a valid claim of continuing violation, see, e.g., McKenzie v. Sawyer, 684 F.2d 62, 72 (D.C.Cir.1982), Milton v. Weinberger, 645 F.2d 1070, 1074-77 (D.C.Cir.1981), Title VII does not apply retroactively to discrimination against federal employees that occurred prior to March 24, 1972, unless a complaint was pending on its effective date. Stoller v. Marsh, 682 F.2d 971, 974 (D.C.Cir. 1982); Thompson v. Sawyer, 678 F.2d 257, 289 (D.C.Cir.1982); Brown v. Turner, 659 F.2d 1199, 1201-02 (D.C.Cir.1981). In this case, the administrative complaint on which the class allegations rest was not filed until June 21, 1972. Under the Civil Service Regulations in effect at that time, all claims originating more than fifteen days prior to that complaint are time-barred. 5 C.F.R. § 713.214(a)(l)(i) (1972) (currently codified at 29 C.F.R. § 1613.214(a)(l)(i) (1982)); see Milton, 645 F.2d at 1072 & n. 6. Although the District Court recognized these fundamental principles, the appellants believe that it erred in refusing to apply them when analyzing the plaintiffs’ statistical evidence.
The District Court concluded that plaintiffs need not factor out time-barred discrimination because defendants may be held liable for the continuing effects of that discrimination. This theory was once in vogue, but it is flatly inconsistent with the Supreme Court’s pronouncements in Hazel-wood School District v. United States, 433 U.S. at 309-10, 97 S.Ct. at 2742-43, and United Air Lines, Inc. v. Evans, 431 U.S. 553, 558, 97 S.Ct. 1885, 1889, 52 L.Ed.2d 571 (1977). As noted above, however, we believe that the District Court’s conclusion that the plaintiffs’ statistics were capable of establishing a prima facie case was correct, albeit in part for the wrong reason, and should thus be affirmed. See Helvering v. Gowran, 302 U.S. 238, 245, 58 S.Ct. 154, 157, 82 L.Ed. 224 (1937); Eikenberry v. Callahan, 653 F.2d 632, 636 (D.C.Cir.1981).
By grouping nonactionable hiring decisions with those for which NAVCOSSACT/NARDAC could properly be held liable, certain of the plaintiffs’ statistical analyses hold the potential for some distortion concerning the adverse impact of the appellants’ post-1972 promotion policies. However, it is clear that the plaintiffs did not rely solely on data regarding nonactionable hiring decisions; indeed, there is no doubt whatsoever that the plaintiffs’ statistical analyses included a substantial amount of data concerning post-1972 employment decisions affecting class members. For these reasons, we do not believe that “failure to factor out time-barred discrimination discredited the analyses. Statistics tuned to the proper time period are more probative than statistics not so tuned, but categorical rejection of the latter is not warranted.” Valentino v. United States Postal Service, 674 F.2d at 71 n. 26. The approach to statistics adopted by the Supreme Court in Hazelwood, 433 U.S. at 309-10, 97 S.Ct. at 2742-43, and International Brotherhood of Teamsters v. United States, 431 U.S. at 360, 97 S.Ct. at 1867, plainly suggests that statistics including time-barred decisions are “sufficient to support a prima facie case and shift the burden to the defendant to show that its actions during the relevant period rebut the inference of discrimination raised by the plaintiff’s broad summary.” Movement for Opportunity & Equality v. General Motors Corp., 622 F.2d 1235, 1258 (7th Cir.1980).12
Although the appellants did attempt to make such a showing, the trial court found their statistical evidence defective on other grounds. We can find no basis upon which to overturn Judge Greene’s findings on this point. Therefore, we agree that the appellants’ evidence was incapable of raising a genuine issue concerning the validity of the inference of discrimination raised by the plaintiffs’ statistical presentation. As a result, NAVCOSSACT/NARDAC was left with only an unquantified theoretical objection that could not undermine the prima facie case. See II Jt.App. 743 — 44.
*3683. Relevant Employer
The appellants’ final objection to the sufficiency of the prima facie case is that the plaintiffs’ statistics attribute to NAVCOSSACT/NARDAC responsibility for the employment decisions of other agencies over which it had no control. Undisputed trial testimony reveals, they claim, that initial grade placements were made by either the CSC or the OPM or by employing agencies that were not made defendants in this case. As we read the record, the District Court’s conclusion that NAVCOSSACT/NARDAC “did not make a convincing showing that it had no control over initial grade determinations,” Trout, 517 F.Supp. at 880, was clearly erroneous. The rebuttal evidence submitted by the defendants at trial was sufficient to raise a genuine issue concerning their responsibility for initial placement decisions, and the plaintiffs made no attempt to counter this evidence. We conclude, therefore, that the District Court erred in holding that the class was entitled to relief for discriminatory initial placements.
This does not mean, however, that the plaintiffs’ statistics could not make out a legally sufficient prima facie case of discrimination in promotions. Because NAVCOSSACT/NARDAC did not quantify the extent to which the disparity between men and women revealed by the plaintiffs’ statistics was attributable to the locked-in effects of nonactionable discrimination, see, e.g., IV Jt.App. 1829, the plaintiffs’ statistical analyses created a justifiable inference that the defendants had “failed to promote equitably individuals who were discriminated against at hiring,” Trout, 517 F.Supp. at 885, or had slowed the progress of women who had initially been properly placed.
C. The Rebuttal Evidence
NAVCOSSACT/NARDAC attempted to bolster its objections to the plaintiffs’ statistical evidence by introducing several statistical studies of its own at trial and by proffering additional analyses at the relief and reconsideration stages. The District Court, however, found the former studies unpersuasive for a variety of reasons and refused to consider the latter as untimely tendered. The appellants object to both of these decisions.
1. The Evidence Offered at Trial
The evidence introduced at trial in opposition to the plaintiffs’ statistical studies consisted primarily of a “cohort” analysis (which examined the flow through NAVCOSSACT/NARDAC’s promotional system of male and female employees with similar hiring dates who started at the same grade level), a promotion study, and several regression analyses. As we have already observed, the District Court’s decision to reject this evidence can be upset only if clearly erroneous. See Medina v. Reinhardt, 686 F.2d at 1007; Payne v. Travenol Laboratories, Inc., 673 F.2d 798, 823 (5th Cir.), cert. denied, - U.S. -, 103 S.Ct. 451, 74 L.Ed.2d 605 (1982). Although the appellants characterize the reasons underlying Judge Greene’s rejection of their studies as unproven speculation, our review of the record has not left us with a “definite and firm conviction that a mistake has been committed.” United States v. United States Gypsum Co., 333 U.S. at 395, 68 S.Ct. at 542.
As the District Court observed, NAVCOSSACT/NARDAC’s regression studies were, for the most part, not introduced into evidence at trial. The defendants relied instead on brief descriptions of the conclusions allegedly supported by studies that were not in evidence. See, e.g., II Jt.App. 761. Judge Greene’s decision to assign little weight to these studies, which, in any event, did not correct the specification errors allegedly afflicting the plaintiffs’ regressions, was totally reasonable. The defendants’ promotion analysis, moreover, consisted solely of twenty-three pages of statistics, see II Jt.App. 636-41, 719-35, which, as far as we can tell, were never adequately explained at trial. Nothing in the record leads us to believe that the District Court’s characterization of this study as less persuasive than the plaintiffs’ regressions was clearly erroneous.
*369The centerpiece of the defendants’ rebuttal presentation was their cohort analysis. The experts who performed this study “divided ... [NAVCOSSACT/NARDAC’s] work force into groups by the year in which the employees joined the agency and the GS grade at which they entered” and examined the “promotion experiences of each of these groups or ‘cohorts’ over time ... to compare the relative progress of men and women.” Trout, 517 F.Supp. at 884. Having reviewed the record on this point, we cannot say that Judge Greene’s ultimate conclusion that the cohort analysis was less persuasive than the plaintiffs’ regressions was clearly erroneous. As the defendants’ expert conceded in another context, see IV Jt.App. 1874, the division of the work force into extremely small segments .made it unlikely that this cohort study would detect sex-based disparities. See Schmid v. Frosch, 680 F.2d 248, 249-50 & n. 4 (D.C. Cir.1982) (per curiam); Woodard v. Lehman, 530 F.Supp. 139, 145 (D.S.C.1982); Segar v. Civiletti, 508 F.Supp. at 698. The District Court could properly conclude, moreover, that the exclusion of a substantial number of employees from the cohort analysis reduced its probative force. And, even though the defendants could not be held liable for discriminatory initial grade' placements, see Part II.B.3. supra, the lack of verification of the critical assumption of proper initial placements substantially reduced the persuasive appeal of the cohort study. See Trout, 517 F.Supp. at 885 & n. 45; Segar, 508 F.Supp. at 698.
2. The Appellants’ Attempt to Retry the Case
The appellants’ objection to the District Court’s disposition of their motion for reconsideration is both meritless and outrageous for what it suggests. At some point litigation must come to an end, even though it is always possible to offer more evidence. Notwithstanding NAVCOS-SACT/NARDAC’s contention that the new evidence might alter the previous finding of class liability, the District Court could properly conclude that the time for rebuttal had come and gone. Although the court had inherent power to reconsider its interlocutory orders concerning class-wide liability, see Laffey v. Northwest Airlines, Inc., 642 F.2d 578, 583-84 (D.C.Cir.1980), the critical question is whether there was good cause to do so. Absent a convincing showing that the District Court’s answer to this question constituted an abuse of discretion, we will not second-guess its decision. Cf. GAF Corp. v. Transamerica Insurance Co., 665 F.2d 364, 370-71 (D.C.Cir.1981) (discussing Rule 60(b) motion for relief from judgment); Washington Mobilization Committee v. Jefferson, 617 F.2d 848, 850 (D.C.Cir.1980) (discussing Rule 59(a) motion to reopen record).
The facts of this case afford no support for a suggestion that the District Court abused its discretion in refusing to reconsider the question of class-wide liability. As Judge Greene properly concluded, the evidence on which the request for reconsideration was based could have been discovered and presented at trial by a duly diligent defendant. Cf. 6A J. Moore, Moore’s Federal Practice ¶ 59.08[3] (1982) (discussing motion for new trial). The defendants had over seven years in which to collect evidence and prepare their case, and they have offered no good reason whatsoever to explain why their statistical analyses were not produced at trial. Even if these analyses might have adequately rebutted the plaintiffs’ prima facie case — an issue on which our comments would be wholly gratuitous— the District Court plainly did not abuse its discretion in concluding that it need not conduct another trial on the question of liability.
We would add that we find it extremely troublesome — in light of the long and complex history of this litigation and in light of Judge Greene’s patient and thoughtful treatment of the case — that the appellants would even propose that the trial court reopen and retry the matter. In the context of this case setting, such an adversarial tactic is irresponsible, insensitive to the extraordinary burdens faced by district courts already overloaded with heavy dockets, and wasteful of precious resources of litigants and the judiciary. Viewed in this posture, *370we consider the appellants’ request for reconsideration an affront to the judicial system.
D. The Clear and Convincing Evidence Standard
Once a class has established liability in a disparate treatment case, “a discrete remedy phase begins. Any individual class claimant may raise a presumption that he is entitled to relief upon a prima facie showing of class membership. The burden of proof then shifts to the employer to rebut that presumption in each individual case.” Milton v. Weinberger, 696 F.2d 94, at 98 (D.C.Cir.1982) (as modified, Dec. 20, 1982). This burden, the District Court held, is quite substantial: After an individual employee lists the assignments or opportunities denied her or describes other ways in which she was victimized, an employer must rebut the presumption of discrimination with “clear and convincing evidence.” See Stewart v. General Motors Corp., 542 F.2d 445, 453 (7th Cir.1976), cert. denied, 433 U.S. 919, 97 S.Ct. 2995, 53 L.Ed.2d 1105 (1977); Baxter v. Savannah Sugar Refining Corp., 495 F.2d 437, 444-45 (5th Cir.), cert. denied, 419 U.S. 1033, 95 S.Ct. 515, 42 L.Ed.2d 308 (1974).
Relying on International Brotherhood of Teamsters v. United States, 431 U.S. at 360-62, 97 S.Ct. at 1867-68, and Texas Department of Community Affairs v. Burdine, 450 U.S. at 254-56, 101 S.Ct. at 1094-95, the appellants argue that the effect of a finding of class-wide discrimination is identical to the effect of a prima facie case in an individual suit. They claim, therefore, that they can defeat individual class members’ entitlement to relief merely by advancing legitimate, nondiscriminatory reasons for their employment decisions. As the appellants appeared to recognize in the oral argument before this court, however, their precise claim was recently considered and rejected in McKenzie v. Sawyer, 684 F.2d at 76-78. We held in McKenzie that the situation of an individual class member after a showing of class-wide liability “is not analogous to the situation of an individual plaintiff who has made out a prima facie case, of disparate treatment.” Id. at 77. Because such class members are presumptively entitled to relief upon showing that they were potential victims of the defendants’ discriminatory practices, Teamsters, 431 U.S. at 362, 97 S.Ct. at 1868, the defendants must “rebut the plaintiffs’ individual showings by clear and convincing evidence.” McKenzie, 684 F.2d at 78. Cf. Day v. Mathews, 530 F.2d 1083, 1085-86 (D.C.Cir.1976) (per curiam) (individual case).
E. The Individual Claims
The appellants’ final challenges to the District Court’s decisions concern its findings of discrimination against two individual plaintiffs, Marie Bach and Clara Perlingiero, and the relief awarded to those individuals. We have found no basis for challenging the District Court’s conclusion that appellee Bach was relieved of her responsibility for automatic data processing security in March 1976 and May 1978 for discriminatory reasons. When tested against the clearly erroneous standard, the court’s conclusions that questions concerning her technical background were not the predominant reason for the personnel actions and that the reasons given were mere pretext for discrimination are impregnable.
When her responsibilities were removed, Bach was a GS-13 security specialist; the duties that she had performed were transferred to males at the GS-14 and GS-15 levels in 1976 and to males at the GS-15 level in 1978. Apparently for this reason, the District Court ordered her retroactively promoted to GS-14 as of March 1976 and to GS-15 as of May 1978. We agree with the District Court that these reassignments could constitute evidence that the defendants had failed appropriately to classify Bach’s position. Trout v. Hidalgo, No. 73-55, mem. op. at 10 n. 12 (D.D.C. Oct. 20, 1981), reprinted in I Jt.App. 16, 25. We believe, however, that further consideration of the propriety of the retroactive promotions is necessary because it is not clear that Bach was ever denied a promotion to GS-14 or GS-15 and the males who assumed her *371duties may have had additional responsibilities that justified their higher classifications.
We have found no basis for upsetting the District Court’s conclusions that “the various actions taken against [Perlingiero] were based on sex discrimination and on retaliation, as distinguished from objective efficiency-related factors,” Trout, 517 F.Supp. at 891, and that she should be awarded a retroactive promotion to GS-14. The promotion to GS-13 made effective September 1, 1971 poses a more troubling question. This award was improper, the appellants argue, because Perlingiero had no administrative complaint pending on March 24, 1972, the effective date of Title VII. Although the record suggests that Perlingiero may have withdrawn a complaint filed on September 1, 1971 in response to promises that her grievance would be remedied, the District Court made no finding to that effect, and we are thus unable to discern the rationale underlying its retroactive promotion to GS-13.
III. Conclusion
For the reasons set forth above, we affirm in part, reverse in part, and vacate and remand in part. The District Court’s findings of class-wide discrimination in promotions and discrimination against Bach and Perlingiero are affirmed, as is its decision imposing the clear and convincing standard of proof in individual relief hearings. The court’s imposition of liability for discrimination in initial grade placements is reversed, and its awards of retroactive promotions to GS-14 and GS-15 and commensurate back-pay to Bach and a retroactive promotion to GS — 13 and commensurate backpay to Perlingiero are vacated and remanded. All other elements of the individual relief are affirmed.
So Ordered.
. Title VII of the Civil Rights Act of 1964, 42 U.S.C. §§ 2000e to 2000e-17 (1976 & Supp. IV 1980).
. Trout v. Hidalgo, 517 F.Supp. 873 (D.D.C. 1981).
. Trout v. Hidalgo, Civ. No. 73-55 (D.D.C. Oct. 20, 1981), reprinted in I Joint Appendix (“Jt. App.”) 16.
. Trout v. Lehman, Civ. No. 73-55 (D.D.C. Mar. 10, 1982), reprinted in I Jt.App. 28.
. Although a more thorough analysis of the evidence offered by the appellants in support of their motion for reconsideration might have led the trial court to conclude that the prima facie case had been rebutted — a proposition about which we will not speculate since the appellants’ evidence was not subjected to the rigors of the adversarial process — we find no basis for overturning the court’s discretionary refusal to retry the case many months after rendering its decision.
. Trout v. Middendorf, Civ. No. 73-55 (D.D.C. June 18, 1979) (order limiting class definition), Record (“R.”) 115; see Trout v. Hidalgo, 517 F.Supp. 873, 877 (D.D.C.1981).
. The District Court also found persuasive the testimony of a number of class members. Although we do not specifically discuss this testimony, the District Court properly considered the statistical and nonstatistical evidence on a cumulative basis in assessing the strength of the plaintiffs’ prima facie case. EEOC v. American Nat’l Bank, 652 F.2d 1176, 1188-89 (4th Cir.1981), cert. denied, - U.S. -, 103 S.Ct. 235, 74 L.Ed.2d 186 (1982).
. Trout v. Lehman, Civ. No. 73-55, mem. op. at 2 (D.D.C. Mar. 10, 1982), reprinted in I Jt.App. 28, 29.
. Trout v. Hidalgo, Civ. No. 73-55, mem. op. at 4 (D.D.C. Oct. 20, 1981), reprinted in I Jt.App. 16, 19.
. Valentino’s holding is limited, we believe, by the fact that the plaintiffs’ expert there had failed to submit qualification data that were available. Id. at 71 n. 23. If data on a particular minimum objective qualification are simply unavailable, Valentino’s requirement may be eased, for “[ejxactness is not required at the prima facie stage.” Medina v. Reinhardt, 686 F.2d at 1008-09 & n. 6.
. The appellants also argue that the plaintiffs’ statistics overstate the qualifications of women in ways that, while not related to the minimum objective qualifications for NAVCOSSACT/NARDAC positions, exaggerate the expected salaries of female employees and reduce the explanatory power of the plaintiffs’ model. They contend, for example, that the plaintiffs’ failure to account for type of education biased their statistics because most computer science degrees are earned by men, and that the plaintiffs’ nongovernmental experience proxy biased their statistics because women, on the average, spend more time out of the labor force than do men. Because the testimony offered at trial did not quantify these objections or show that such differences existed in the population of NAVCOSSACT/NARDAC employees, the District Court could, at most, speculate that the omissions might have had some effect on the statistical showing. See, e.g., II Jt.App. 758-60; IV Jt.App. 1880-85. Such speculation, based on evidence not “keyed to the population here involved,” Trout, 517 F.Supp. at 881, would do little to undermine the plaintiffs’ prima facie case. See Vuyanich v. Republic Nat’l Bank, 505 F.Supp. at 307.
. See EEOC v. American Nat’l Bank, 652 F.2d at 1188, 1194-95; B. Schlei & P. Grossman, Employment Discrimination Law 326-27 (Supp. 1979). But see EEOC v. United Va. Bank/Seaboard Nat’l, 615 F.2d 147, 150-51 (4th Cir.1980.)