Judge CALABRESI concurs in the opinion of the Court, and also files a separate opinion.
FEINBERG, Circuit Judge:Plaintiff Arthur Hollander appeals from the grant of summary judgment in the United States District Court for the District of Connecticut, Warren W. Eginton, J., in favor of defendant American Cyan-amid Company. As of this writing, Hollander’s case has been in the federal courts for more than 13 years. This is also the second time that this court has heard the matter, having remanded the case more than eight years ago in order to allow Hollander to conduct additional limited discovery. Hollander v. American Cyanamid Co., 895 F.2d 80 (2d Cir.1990). Hollander appeals principally from the dismissal of his action asserting a claim for age discrimination in violation of the Age Discrimination in Employment Act (ADEA), 29 U.S.C. § 621 et. seq. He also appeals from the partial grant of Cyanam-id’s motion to strike various portions of his affidavit submitted in support of his memorandum in opposition to Cyanamid’s motion for summary judgment. The district court dismissed the suit on the ground that Hollander’s evidence was insufficient to create an inference that Cyanamid discriminated against him because of his age. See Hollander v. American Cyanamid Co., 999 F.Supp. 252 (D.Conn.1998). For the reasons stated below, we affirm.
I. Background
Hollander began his career with Cyan-amid in 1956 as an employee at the company’s Wallingford, Connecticut facility. He was 30 years old. He left the company in 1960, and returned in 1973 to assume the position of Manager of Management Engineering at Cyanamid’s Lederle facility in Pearl River, New York. At the time of his rehire, he was 47.
As Manager of Management Engineering, Hollander received a yearly progress review from his supervisors that rated his ability in several specific areas. The reviews also included comments that elaborated upon or supplemented the ratings. Hollander’s reviews show that Cyanamid was consistently impressed with his creativity and commitment to his job. He also received regular salary increases, bonuses and stock options. However, the progress reviews, which were conducted by four different supervisors over the course of Hollander’s re-employment at Cyanamid, also consistently documented the perception that Hollander has an abrasive man*196agement style and a difficult personality (hereinafter referred to collectively as “interpersonal problems”). Hollander’s 1974 review noted that he had “created some antagonism by ‘shooting from the hip.’ ” His 1976 review stated that “[improvement is needed in establishing a higher confidence with middle management and communicating on a level which they understand.” Hollander’s 1977 review rated him as “needs improvement,” the lowest possible rating, in the category of “communication,” and noted that Hollander needed to “be more understanding of ... managers’ problems and less antagonistic.” In 1978, he was rated as “needs improvement” in both “communication” and “human relations.” The review commented that Hollander needed improvement “in following communications lines” and that “[t]he tendency to go over anyone who does not immediately respond must be changed.” While Hollander’s January 1980 review rated him “consistently acceptable” in the categories of “communication” and “human relations,” the perceived problems returned in his December 1980 review, in which he was once again rated as “needs improvement” in those categories.
In 1981, Cyanamid transferred Hollander to its Davis & Geek facility in Danbury, Connecticut, and appointed him Manager of Medical Devices. The change of setting, however, did not alter the pattern established in Hollander’s earlier reviews. His 1981 review rated him as “needs improvement” in “human relations.”' In 1982, he received a “needs improvement” rating in “human relations,” “safety,” and “maturity & judgment.” The review noted that “several projects are not getting proper attention because of interpersonal relationship problems.”
Hollander received his worst review in 1983, in which supervisor Robert Duckett rated him as “needs improvement” in the areas of “communication,” “human relations,” “personnel development,” “leadership” and “organizing/staffing.” Duckett elaborated:
[Hollander’s] major problem is a total lack of comprehension of the vital necessity of properly considering all of the appropriate functional groups which must be involved for successful completion and execution of complex projects. He leaves a “trail of wreckage” of interpersonal and inter departmental relationships which eventually inhibit the success of the projects. He seems to ignore the vital “people function” almost completely, and doesn’t comprehend this effect. He operates like a “one man gang” which is not acceptable in a complex, multi-functional team oriented business.
In addition, Hollander’s overall job performance for that year was rated as “needs improvement,” the second-worst possible overall rating.1
Cyanamid terminated Hollander in January 1984. At the time, he was 58 years old.2 He was given six months of severance pay and was told that his management profile would be circulated within Cyanamid to determine if a different division was interested in his services. Hollander was discharged permanently in August 1984. Just before that — in July 1984 — he filed age discrimination complaints with the Connecticut Commission on Human Rights and Opportunities (CCHRO) and the Equal Opportunity Employment Commission (EEOC).
In October 1984, Hollander sought employment with Ethicon, Inc., a medical device manufacturer and competitor of Cyan-amid. Hollander offered to show Ethicon a Cyanamid film demonstrating the appli*197cation of automation to suture manufacturing, which led Ethicon to contact Cyanam-id regarding both Hollander and the film. Cyanamid wrote Hollander, stating that his employment at Ethicon would be inconsistent with a non-competition agreement he had signed in 1973, and that his offer to show Ethicon the film infringed Cyanam-id’s proprietary rights. Without mentioning the non-competition agreement or the dispute over the film, Ethicon informed Hollander that it would not be offering him a position.
In August 1985, Hollander filed suit in the United States District Court for the District of Connecticut, claiming that Cyanamid discharged him in violation of the ADEA. He also claimed that- Cyanam-id’s opposition to his employment at Ethi-con was retaliation for the initial complaints he had brought before the CCHRO and the EEOC and, as such, was a second violation of the ADEA. Hollander further contended that Cyanamid’s alleged retaliation amounted to tortious interference with a business expectancy under Connecticut law. Finally, he alleged that Cyanamid’s practices violated federal and state antitrust laws and state law prohibitions on unfair trade practices. Hollander later dropped the antitrust and unfair trade practices claims.
In March 1989, the district court granted Cyanamid’s motion for summary judgment on Hollander’s ADEA claims. The court held that Hollander had failed to show a nexus between his age and his discharge, and had failed to link Cyanam-id’s conduct regarding Ethicon to Hollander’s filing of discrimination charges with the CCHRO and the EEOC. Because Hollander’s federal claims had been disposed of by summary judgment, the district court also dismissed Hollander’s remaining state law tortious interference claim for lack of jurisdiction.
On appeal to this court, the judgment of the district court was vacated in part, affirmed in part and remanded. 895 F.2d 80. The panel affirmed the district court’s judgment as to Hollander’s retaliation claim. Id. at 85-86. However, the panel also held that the district court had erred when it refused to compel Cyanamid to answer an interrogatory asking it to identify each management-level employee over the age of 40 whose employment with Cyanamid had been terminated since 1983. Id. at 85. The panel accordingly remanded the case, deeming the additional discovery necessary to allow Hollander the opportunity to gather statistical evidence which “might uncover a pattern of older management employees leaving ... Cyan-amid under unexplained circumstances, which might help prove his claim that ... Cyanamid’s explanation for his discharge was pretextual.” Id. at 84. The panel noted that “if following the completion of discovery, the district court finds that Hollander has not established a genuine issue of material fact suggesting that ... Cyan-amid’s discharge of Hollander was pretex-tual, it may enter summary judgment for ... Cyanamid.” Id. at 85.
After remand, Hollander expanded his discovery efforts to include several additional interrogatories and requests for production other than the single interrogatory at issue on the prior appeal. When Cyan-amid did not respond to the additional requests, Hollander filed a Motion to Compel and for Sanctions. Recommending denial of the motion in November 1994, Magistrate Judge Thomas P. Smith noted that Hollander had “received discovery of an extent unprecedented in this magistrate’s experience of over 15 years,” and “well beyond the contemplation of the Second Circuit’s decision, which the magistrate has interpreted with great fairness toward [him].” The district court adopted the recommendation.
In April 1997, Cyanamid moved for summary judgment on Hollander’s claims and moved to strike, as inconsistent with Fed. R.Civ.P. 56(e), certain portions of the affidavit Hollander had submitted in support of his memorandum in opposition to summary judgment. Cyanamid’s motion to *198strike was granted in part and denied in part; the district court struck numerous portions of Hollander’s affidavit that, according to the court, contained “inadmissible hearsay, generalized conclusory and argumentative statements, and/or information which could not be attributed to plaintiffs personal knowledge.” 999 F.Supp. at 256. The court did not strike everything that Cyanamid argued was inappropriate and, insofar as some of the stricken portions quoted or paraphrased other evidence that was properly before the court, the court separately considered that evidence. Id. The court granted Cyanamid’s motion for summary judgment on the ADEA claim, holding that Hollander’s evidence did not create an inference that Cyanamid discriminated against him because of his age. Id. at 260-61. Having again dismissed Hollander’s federal claim, the court also again dismissed his state claim for tortious interference for lack of jurisdiction. Id. at 261. This appeal followed.
II. Discussion
A. Motion to Strike Portions of Hollander’s Affidavit
Rule 56(e) states that:
[supporting and opposing affidavits shall be made on personal knowledge, shall set forth such facts as would be admissible in evidence, and shall show affirmatively that the affiant is competent to testify to the matters stated therein.
A court may therefore strike portions of an affidavit that are not based upon the affiant’s personal knowledge, contain inadmissible hearsay or make generalized and conclusory statements. See, e.g., United States v. Private Sanitation Industry Ass’n of Nassau/Suffolk, Inc., 44 F.3d 1082, 1084 (2d Cir.1995). Hollander argues that the district court erred in striking portions of his affidavit because the stricken provisions were (1) harmless summaries or conclusions incorporating discovery materials contained elsewhere in the record, (2) Hollander’s first-hand knowledge of his own work, Cyanamid’s records and practices, and the alleged personal knowledge of Cyanamid’s witnesses, and (3) Hollander’s description of the statistical evidence.
We will not disturb a district court’s grant of a motion to strike unless manifestly erroneous. Cf. Luciano v. Olsten Corp., 110 F.3d 210, 217 (2d Cir.1997). Hollander correctly notes that some of the stricken provisions were mere summaries of evidence properly included elsewhere in the record. That, however, does not lead to the conclusion Hollander desires because, as already noted, that evidence was separately considered by the district court. Hollander, 999 F.Supp. at 256. Hollander’s remaining contentions are without merit. The district court was correct in holding that the affidavit “flagrantly disregarded] the requirements of Rule 56(e),” was “riddled with inadmissible hearsay, conclusory statements and arguments, and information clearly not made on the affi-ant’s personal knowledge,” and “more re-semblefd] an adversarial memorandum than a bona fide affidavit.” Id. The court acted well within its discretion by striking the inappropriate portions of the affidavit.
B. Hollander’s ADEA Claim
The ADEA forbids employers from discriminating in hiring, discharge, or the setting of “compensation, terms, conditions, or privileges of employment” based upon the age of an employee. 29 U.S.C. § 623(a)(1); see also Hazen Paper Co. v. Biggins, 507 U.S. 604, 609, 113 S.Ct. 1701, 123 L.Ed.2d 338 (1993). The ADEA covers the class of individuals who? like Hollander, are over the age of 40, 29 U.S.C. § 631(a), and the evidentiary framework for proving age discrimination under the statute is the same as that for proving discrimination under Title VII. See Raskin v. Wyatt Co., 125 F.3d 55, 60 (2d Cir.1997). Thus, in a disparate treatment case like this one we follow the three-step, *199burden-shifting analysis set forth in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973) and Texas Dep’t of Community Affairs v. Burdine, 450 U.S. 248, 101 S.Ct. 1089, 67 L.Ed.2d 207 (1981). Briefly, the analysis is as follows. First, the plaintiff must establish a prima facie case of discrimination. The burden then shifts to the employer to rebut the prima facie case by providing a legitimate, non-discriminatory reason for its actions. The plaintiff then has the opportunity to show that the reason offered by the employer was not its true reason, and that age was. The question that must be asked is whether the plaintiff has shown, by a preponderance of the evidence, that he has been the victim of age discrimination. Thus, the “ultimate burden of persuading the trier of fact that the defendant intentionally discriminated against the plaintiff remains at all times with the plaintiff.” Burdine, 450 U.S. at 253, 101 S.Ct. 1089 (citations omitted) (emphasis supplied).
1. Hollander’s Prima Facie Case
To establish a prima facie case of discrimination, Hollander had to show that (1) he was within the protected age group, (2) he was qualified for the job, (3) he was discharged, and (4) the discharge occurred under circumstances giving rise to an inference of age discrimination. See Hollander, 895 F.2d at 83. The burden of proof that must be met to establish a prima facie case is minimal. See Chambers v. TRM Copy Ctrs. Corp., 43 F.3d 29, 37 (2d Cir.1994); Dister v. Continental Group, Inc. 859 F.2d 1108, 1114 (2d Cir.1988). This court held in Hollander’s first appeal that he had established a prima facie case for purposes of summary judgment: he was 58 years old when discharged, was qualified to perform his job as Manager of Medical Devices, and his duties were allegedly assigned to two replacements, the first of whom (Richard Augsbach) was approximately eight months younger and the second (Felix Es-posito) 11 years younger.3 See Hollander, 895 F.2d at 83.
2. Cyanamid’s Articulation of a Legitimate, Non-Discriminatory Reason
Since Hollander established a prima fa-cie case, the burden shifted to Cyanamid to articulate a legitimate, non-discriminatory reason for his discharge. Cyanamid points to the interpersonal problems that were consistently documented in Hollander’s personnel reviews. Cyanamid also claims that Hollander had to be removed from a project because of his poor interpersonal skills, and that his strained relations with certain employees he supervised caused them to seek union representation.
Hollander’s brief attempts to show that Cyanamid failed to put forth a legitimaté, non-discriminatory reason for his discharge. Hollander argues that the district court erred in failing to strike as hearsay and violative of the best evidence rule the deposition testimony of various Cyanamid employees, including Davis & Geek Director of Operations Gill Seal, Personnel Director John O’Shaughnessy and Personnel Department employee Robert D’Andrea, which supported Cyanamid’s stated reason. This contention, however, *200does not merit much discussion. First, the district court’s opinion after remand did not refer to the challenged statements at all and did not appear to consider or rely on them in concluding that Cyanamid sufficiently articulated a legitimate, nondiscriminatory reason for Hollander’s discharge. Further, this court held in Hollander’s first appeal that Cyanamid had satisfied its burden of producing a non-diseriminatory motive for Hollander’s termination. See Hollander, 895 F.2d at 83. Finally, even if the district court did err in failing to strike these statements, there was ample other evidence, documented in year after year of progress reviews, from which to conclude that Cyanamid had furnished a legitimate reason with adequate specificity. Accordingly, the district court’s error, if any, in refusing to strike the challenged statements was harmless.
3. Hollander’s Attempt to Show that he is a Victim of Age Discrimination
Cyanamid’s articulation of a legitimate, non-diseriminatory reason brings us to what the district court called “the heart of this case” — whether Hollander presented sufficient evidence for a reasonable jury to conclude that Cyanamid discriminated against him because of his age. 999 F.Supp. at 257. Hollander put forth two categories of evidence in support of his position. First, he offered evidence to rebut Cyanamid’s non-discriminatory reason by raising a question of fact as to whether he actually had interpersonal problems; in other words, Hollander attempted to show that Cyanamid’s stated reason for his discharge was pretextual and itself constituted proof of age discrimination. Second, he offered statistical evidence from which, he argued, a reasonable jury could infer that Cyanamid discriminated against him because of his age. While the district court found that the evidence submitted by Hollander “may in fact raise some doubt as to whether [Cyanamid’s] articulated reason for discharging him is believable”, 999 F.Supp. at 258, it held that no reasonable jury could conclude that, based upon that evidence, Hollander was a victim of age discrimination. For the reasons stated below, we agree.
a. Hollander’s non-statistical evidence
Various items made up the first category of evidence on which Hollander relies. Hollander offered affidavits from former co-workers stating that he was not difficult to work with and may not have been responsible for the alleged unionization incident cited by Cyanamid. Hollander also stressed the fact that, at his depositions in 1986 and 1987, Duckett could not remember a single event that led to the scathing 1983 review of Hollander in which Duckett wrote that Hollander left a “trail of wreckage” in regard to interpersonal relationships. In addition, Hollander offered letters from employees at Cyanamid — including Duckett — complimenting his efforts as Manager of Medical Devices.
We need not decide whether Hollander offered enough evidence to establish that Cyanamid used his allegedly poor interpersonal skills as a pretext for its true reason for terminating him. Assuming arguendo that he did, this is far from the end of the matter, for to survive summary judgment Hollander had to show not only pretext, but also either use of a pretext that itself implies a discriminatory stereotype, or use of a pretext to hide age discrimination. This, he did not do.
This court was presented with a similar situation in Fisher v. Vassar College, 114 F.3d 1332 (2d Cir.1997) (en banc) (Fisher III), cert. denied, — U.S. -, 118 S.Ct. 851, 139 L.Ed.2d 752 (1998). In Fisher III, plaintiff Cynthia Fisher asserted claims of sex and age discrimination after being denied tenure by her former employer, Vassar College. Vassar argued that Fisher was denied tenure because of a review that found her scholarship, teaching ability, leadership and service to the college to be deficient. In an attempt to show that Vassar’s reason was pretextual, Fisher offered evidence that contradicted *201certain conclusions of the review. The district court, after a bench trial, concluded in Fisher v. Vassar College, 852 F.Supp. 1193 (S.D.N.Y.1994) (Fisher I) that Vassar had unlawfully discriminated against Fisher on the basis of her sex. A panel of this court, however, reversed the district court. Fisher v. Vassar College, 70 F.3d 1420 (2d Cir.1995) (Fisher II). Thereafter, the en banc court in Fisher III reached the same result as the panel did in Fisher II because Fisher’s evidence did not show that Vassar’s stated reason for denial of tenure was a pretext for discrimination 071 the basis of her sex or itself reflected a discriminatory stereotype. Fisher III, 114 F.3d at 1333. The en banc majority opinion stated that a “plaintiff may prevail only if an employer’s proffered reasons are shown to be a pretext for discrimination, either because the pretext finding itself points to discrimination or because other evidence in the record points in that direction — or both.” Id. at 1339; see also Quaratino v. Tiffany & Co., 71 F.3d 58, 64 (2d Cir.1995) (“An employer’s reason for termination cannot be proven to be a pretext for discrimination unless it is shown to be false and that discrimination was the real reason.”). While Fisher had produced some evidence of pretext, neither the pretext nor the record pointed to the conclusion that Vassar had discriminated against her on the basis of her sex. Elaborating, the majority of the en banc court stated in Fisher III that
discrimination does not lurk behind every inaccurate statement. Individual decision-makers may intentionally dissemble in order to hide a reason that is non-discriminatory but unbecoming or small-minded, such as back-scratching, log-rolling, horse-trading, institutional politics, envy, nepotism, spite, or personal hostility ... In short, the fact that the proffered reason was false does not necessarily mean that the true motive was the illegal one argued by the plaintiff.
114 F.3d at 1337-38.
The district court in this case reached the same conclusion, and we agree with Cyanamid that the court did not commit reversible error in doing so. First, assuming arguendo that Hollander’s alleged interpersonal problems were used by Cyan-amid as a pretext for the company’s true reason for his termination, the pretext in and of itself does not suggest age discrimination. Had Cyanamid claimed that it terminated Hollander because of a characteristic stereotypically associated with age — for example, because he was “unable to keep up with the times” — and had Hollander shown that reason to be pretextual, our conclusion might be different.4 The perception that Hollander had interpersonal problems — noted almost from the outset of his re-employment by Cyanamid — does not reflect such stereotypical thinking.
Second, nothing else in the record points to age discrimination as the real reason for Hollander’s termination. As we noted in Fisher III, an employer may give a false explanation for terminating an employee in order to mask the true reason which, though petty, spiteful or otherwise ignoble, is not unlawful. Hollander overlooks this statement of the law by contending that if Cyanamid’s preferred reason is not to be believed, then the true reason for his termination must have been his age. However, Hollander put forth no proof of any age-related remarks or actions toward him by Cyanamid personnel, and admitted that his age was never explicitly raised as a factor in his termination. If reliance on Hollander’s alleged interpersonal problems was indeed a pretext — and we repeat that *202we are not so holding — the record before us shows that the stated reason may have been a pretext for various won-discriminatory reasons.5 This is not a case in which a finder of fact could reasonably conclude that Hollander’s termination was, more probably than not, due to age discrimination because no other, non-discriminatory explanation was possible. The district court concluded that “of the many possible reasons for Cyanamid’s inconsistencies, illegal discrimination is no more likely a reason than any other.” Hollander, 999 F.Supp. at 259. We agree. Therefore, it was not error for the district court to hold that this evidence did not permit the inference that Cyanamid discriminated against Hollander because of his age.
b. Hollander’s statistical evidence
We are thus left with the question whether Hollander’s statistical evidence was sufficient to prevent the grant of summary judgment. As this court has held in several cases, disparate treatment plaintiffs may introduce statistics as circumstantial evidence of discrimination. See, e.g., Stratton v. Dep’t for the Aging, 132 F.3d. 869, 877 (2d Cir.1997); Hollander, 895 F.2d at 84. Hollander offered two bodies of such evidence to the district court. The first was a statistical analysis and report prepared by Dr. Marie Gau-dard, Ph.D. (the Gaudard Report), a statistics professor at the University of New Hampshire. Gaudard analyzed the terminations of Cyanamid managerial employees whose jobs were eliminated — that is, employees who were terminated because their positions at Cyanamid ceased to exist and were theoretically never replaced. Gaudard claimed that, to a high degree of statistical certainty, age was a determining factor in the terminations of this “job elim-mated” category of employees. The second body of statistical evidence, which Hollander calls the “title-switching evidence,” was Hollander’s own study of Cyanamid’s personnel directories which, he claims, justifies an inference of age discrimination.
Hollander argues that the district court usurped the role of the jury by weighing the strength of, and ultimately rejecting, both bodies of evidence. Our reading of the district court’s opinion, however, is that the court determined that the evidence was not at all probative of age discrimination against Hollander. A district court’s discretion in choosing whether or not to admit evidence is broad. See United States v. Local 1804-1, Int’l Longshoremen’s Assoc., 44 F.3d 1091, 1095 (2d Cir.1995). Like any other evidence, the opinion of an expert is not immune from the relevance requirement of Rule 401 of the Federal Rules of Evidence. See United States v. Khan, 787 F.2d 28, 34 (2d Cir.1986). Indeed, a district court must act as a gatekeeper to exclude invalid and unreliable expert testimony. See General Elec. Co. v. Joiner, 522 U.S. 136, 118 S.Ct. 512, 517, 139 L.Ed.2d 508 (1997); Raskin, 125 F.3d at 65-66 (rejecting, in ADEA case, plaintiffs argument that district court “usurped the role of the jury” when it held inadmissible expert’s statistical report of employer’s promotion and termination practices). We therefore review the district court’s ruling for abuse of discretion. See Joiner, 118 S.Ct. at 517.
(i). The Gaudard Report
We begin with the Gaudard Report. First, Gaudard plotted, in 10-year groupings, the age distribution of Cyanam-id managers compensated at salary level “9” and higher (level 9 4-) during the period 1984-19896. The Report suggests that *203a five percent decrease in level 9 + managers age 50-59 during that time is evidence that Cyanamid’s terminations did not occur in an age-neutral manner. However, the Report’s inference of discrimination solely on the basis of the raw numbers is impermissible in the absence of any attempt to account for other causes of the age 50-59 anomaly. See Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 657, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989) (requiring a showing of causation between the challenged practice and the alleged disparities); Raskin, 125 F.3d at 67-68. Gaudard made no attempt in this part of the Report to examine how many level 9 + managers age 50-59 left Cyanamid voluntarily from 1984 to 1989, or how many people in that age grouping applied to Cyanamid during that period. Highlighting Gaudard’s failure to account for explanations other than discrimination is the fact that both the age 60-69 and 70-79 level 9+ workforce increased moderately over the course of those years7. In short, the data presented by Gaudard in this part of the Report does not make it any more or less likely that Cyanamid is liable for age discrimination.
The second part of the Report is also flawed although somewhat more sophisticated. Gaudard analyzed the terminations of Cyanamid managers age 40 and over and compensated at salary level “9” and higher (level 9+/age 40+ managers) during the periods 1981-1991, 1981-1984 and 1984-1989. Using numerical termination codes that corresponded to the reason for each termination, Gaudard excluded from her analysis employees who were voluntarily terminated for reasons such as “resignation,” “other employment” and “relocation.” She also excluded involuntary terminations for reasons such as “unsatisfactory performance” and “insubordination.” Finally, she excluded retirements, deaths and group terminations. From the original group of approximately 1300 level 9+/age 40+ managers, Gaudard was left with 245 who had been involuntarily terminated because Cyanamid had eliminated their jobs (the “job eliminated” group). Gaudard made this “job eliminated” group the company-wide statistical sample for her analysis. She claimed that the statistical likelihood of observing as many “job eliminated” terminations of managers ages 55-69 as was observed over the three relevant time periods referred to above “is virtually zero,” and “conclude[d] with a high degree of statistical certainty that ‘job eliminated’ was not an age-neutral termination reason.”
However, the conclusion of the Report that underlies Hollander’s argument here depends on a misleading age grouping that impermissibly skews the data in Hollander’s favor. See Fisher II, 70 F.3d at 1443 (a plaintiff may not gerrymander data in order to achieve a desired result). In contrast to the 10-year age groupings in the rest of the Report, Gaudard’s “job eliminated” analysis uses the 15-year grouping of ages 55-69 to show Cyanam-id’s allegedly discriminatory termination pattern. That grouping, of course, encompasses Hollander’s age at the time he was fired — 58. A close examination of the data, however, shows that it was managers age 60-69 who had their jobs eliminated at a much higher rate than all other managers, including those age 55-59. In other words, it is the job eliminations in the 60-69 segment of the 55-69 age grouping that accounts for the disparity between job eliminations for managers age 40-55 and those age 55-69. Thus, we believe that the district court did not err in concluding that the Gaudard Report was not sufficiently probative of age discrimination against Hollander to defeat summary judgment.
ii. The Title-Switching Evidence
The second body of statistical evidence offered by Hollander was his own *204study of Cyanamid’s personnel directories, the “title-switching” evidence. According to Hollander, the directories show that 26 positions out of the 245 that Cyanamid coded as “job eliminated” still exist. He claims that approximately half of those 26 positions were filled by employees significantly younger than those who were purportedly “job eliminated.” He thus concludes that a number of managers whose terminations were coded as “job eliminated” were actually terminated because of their age and replaced by younger employees.
We agree with the district court that unreliability problems are also present in the title-switching evidence. The district court correctly pointed out that Hollander offered no evidence showing that the so-called younger replacements actually perform the same duties as the terminated managers whom Cyanamid allegedly mis-characterized as “job eliminated.” 999 F.Supp. at 260. By failing to place the text of the directories, which continue to list job titles similar or identical to some of those Cyanamid claims were eliminated, within any sort of factual context, Hollander is unable to show that the younger employees in those positions are truly replacements. The district court also pointed out that “plaintiffs study fails to account for the fact that of those 26 [so-called replacements], roughly half ... were aged 40 or older.” Id. Finally, the district court noted that the evidence does not consider the possibility that even if Cyanamid mischaracterized the 26 positions as “job eliminated,” it terminated the employees who performed those jobs for non-discriminatory reasons. Id.
Accordingly, we hold that the district court acted within its discretion in holding the statistical evidence inadmissible because no reasonable jury could infer from it that Cyanamid discriminated against Hollander because of his age.
Conclusion
We have considered all of Hollander’s arguments, and find that they do not justify reversal. We conclude that (1) the district court did not err in striking portions of Hollander’s affidavit because they were not based on plaintiffs personal knowledge, contained inadmissible hearsay and made conclusory statements; and (2) Hollander’s evidence was insufficient to create an inference that defendant discriminated against him because of his age. The judgment of the district court is, therefore, affirmed.
. The only worse possible overall rating was “consider for termination.”
. Hollander claims that he was 58 in January 1984. The district court said Hollander was 57 when he was terminated. 999 F.Supp. at 254. We do not regard the difference as significant, and use Hollander's figure.
. Hollander filed his complaints with the CCHRO and the EEOC in July 1984, at which time his former job was being performed by the 57-year old Augsbach. Hollander argues that Augsbach was a temporary "floating'' replacement, and that his true replacement was Esposito. Esposito, however, did not even apply for a position at Cyanamid until November 1984, and began working for the company in March 1985. Although the replacement of Hollander with an "insignificantly younger” person like Augsbach would not give rise to an inference of discrimination, see O'Connor v. Consolidated Coin Caterers Corp., 517 U.S. 308, 312-13, 116 S.Ct. 1307, 134 L.Ed.2d 433 (1996) ("In the age-discrimination context, such an inference [that an employment decision was based on age] cannot be drawn from the replacement of one worker with another worker insignificantly younger.”), on a motion for summary judgment, we must construe the facts in the light most favorable to plaintiff.
. As noted by the district court, 999 F.Supp. at 253 n. 1, despite the "copious” amount of discovery in this case, Hollander’s non-statistical evidence attacking the truthfulness of Cyanamid’s reason for his termination is weaker than the evidence of pretext offered by Fisher. Id. at 259 n. 3. Unlike Fisher, Hollander offered insufficient evidence to permit an inference that the supervisors who poorly rated his interpersonal skills actually believed that they were adequate; Hollander’s affidavit evidence shows only the opinions of his coworkers, not those of his supervisors.
. One example of a possible non-discriminatory reason is the apparent friction between Hollander and Duckett. In one incident, Hollander (unknown to Duckett) sent Duckett’s superior a copy of a memorandum Hollander wrote to Duckett disagreeing with Duckett's decision to remove him from a particular project. Duckett was sufficiently displeased by Hollander’s action, which he considered "out of line” and "insubordination,” to meet with Hollander and memorialize it in an internal memorandum.
. We are informed that the data provided to Hollander by Cyanamid permitted analysis for the years 1984, 1985, 1987 and 1989 only.
. It is also worth noting that, during the 1984-1989 period, the greatest increase in Cynamid's level 9+ workforce occurred in the 40-49 age group — employees within the class of persons covered by the ADEA.