OPINION
MILLER, Judge:¶ 1 After a jury trial, Joseph Romero was convicted of second-degree murder. State v. Romero, 236 Ariz. 461, ¶ 1, 341 P.3d 493, 496 (App.2014) (Romero I), vacated in part, 239 Ariz. 6, 365 P.3d 358 (2016) (Romero II). He raised multiple issues on appeal and this court affirmed his conviction. Id. Our supreme court granted Romero’s petition for review on a single issue—whether the trial court abused its discretion by precluding Romero from offering expert testimony criticizing the methods used by firearms examiners to match a gun to a crime. Romero II, 239 Ariz. 6, ¶¶ 1, 10, 365 P.3d at 360, 361. The court held that the trial court erred, vacated a portion of our opinion, and remanded the case to this court for a substantive review of whether the error was harmless. Id. ¶¶ 24, 31. Because we find that the state has not proven beyond a reasonable doubt that the jury would have convicted Romero even had it heard the precluded evidence, we reverse.
Factual and Procedural Background
¶ 2 A detailed review of the facts appears in Romero II, thus we limit our consideration to facts pertaining to harmless error. Id. ¶¶ 2-9. Romero became a suspect in a seven-year-old homicide after detectives, using previously-unexamined information in a cell phone left at the scene, questioned whether a .40-caliber dock Romero had allegedly discarded a month after the homicide could have been used to shoot the victim. Id. ¶¶ 2-5. Police firearms expert Frank Powell test-fired the gun to retrieve the shell casings, which he compared with casings found at the scene of the murder. Id. ¶ 6. He concluded the casings matched. Id. Romero denied involvement in the murder, but did not challenge Powell’s opinion by using testimony from another firearms identification expert. Id. ¶ 6.
¶ 3 After the first trial ended in a mistrial because the jury could not reach a verdict, Romero proffered Ralph Haber as an expert in the field of experimental design. Id. Haber generally opined that forensic firearms identification relies on unscientific standards and methods. Id.; see also Romero I, 236 Ariz. 451, ¶ 12, 341 P.3d at 497. Romero also sought to exclude PoweU’s testimony, arguing that firearms identification did not meet the requirements of Rule 702, Ariz. R. Evid. Romero II, 239 Ariz. 6, ¶ 6, 365 P.3d at 360-61. The state moved to preclude Haber’s testimony and opposed the motion to preclude Powell’s toolmark testimony. Id. ¶¶ 6-7. The trial court denied the motion regarding Powell, but granted the state’s motion, finding that Haber was not qualified as an expert in firearms identification and that his testimony would impermissibly allow the jury to make decisions generally reserved for a Daubert1 hearing. Id. ¶ 7. Our supreme court concluded it was error to preclude Haber’s testimony because he was qualified in scientific experimental design, potential deficiencies in the design of experiments relating to tool-mark analysis were relevant in assessing Powell’s opinions, Haber’s opinion did not impinge on the trial court’s Rule 702 responsibilities, and Haber’s lack of practical experience in toolmark analysis only went to the weight of his testimony. Romero II, 239 Ariz. 6, ¶¶ 17-29, 365 P.3d at 362-64. As noted above, the court remanded the case to this court to determine whether the preclusion of Haber’s testimony was harmless. Id. ¶¶ 30-31.
Discussion
¶4 We requested supplemental briefing from the parties to consider whether, in light of our supreme court’s reasoning, preclusion of Haber’s testimony2 requires us to reverse Romero’s convictions. The parties empha*507sized different aspects of the record. The state primarily adopted the reasoning of onr concurring colleague in Romero I and argued that circumstantial evidence connecting Romero to the scene of the crime was sufficiently persuasive to render it extremely unlikely that the ballistics testimony was incorrect in also linking the gun to the crime, and therefore the preclusion of Haber’s opinion challenging that link was harmless. 236 Ariz. 451, ¶ 69, 341 P.3d at 511 (Eckerstrom, C.J., specially concurring). Romero expanded his original argument to explain how Haber’s testimony would have supplemented the cross-examination of Powell, and he suggests Haber could have convinced the jury to reject Powell’s opinion that the fatal bullets came from the gun linked to Romero.
¶ 5 The state focuses its argument principally on whether the evidence unrelated to firearm identification shows overwhelmingly that the jury would have convicted Romero. This is a guilt-focused argument that generally considers the weight of the untainted, admissible evidence. In contrast, Romero contends that Powell’s testimony provided the bedrock of a guilty verdict. This error-focused argument primarily considers the effect of the error on the trial. To address these arguably separate and independent perspectives, we first examine Arizona principles in oui' black-letter law and then the factors developed in case law.
¶ 6 Although Arizona’s Rules of Civil Procedure and the Federal Rules of Criminal Procedure instruct courts to disregard any trial error that does not affect substantial rights, there is no analogous rule of Arizona criminal procedure. See Ariz. R. Civ. P. 61 (courts must disregard error that does not affect the substantial rights of the parties); Fed. R. Crim. P. 52(a) (error not affecting substantial rights must be disregarded). Arizona constitutional and statutory law, however, proscribe reversal for a trial error if “substantial justice has been done,” Ariz. Const. art. VI, § 27, or there has been no prejudice to a “substantial right” of the defendant, A.R.S. § 13-3987. Additionally, the United States Constitution provides independent protections. See Chapman v. California, 386 U.S. 18, 21, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967) (“[W]e cannot leave to the States the formulation of the authoritative laws, rules, and remedies designed to protect people from infractions by the States of federally guaranteed rights.”). While an evidentiary error does not necessarily rise to the level of a deprivation of a constitutional right,3 the test for whether a substantial right has been affected does not vary depending on whether the error arises out of procedural or constitutional law. See id. at 21-22, 87 S.Ct. 824 (declining to adopt rule that all federal constitutional errors are harmful); see also State v. Ring, 204 Ariz. 534, ¶ 45, 65 P.3d 915, 933 (2003) (noting constitutional error may be harmless).
¶ 7 “In deciding whether error is harmless, the question ‘is not whether, in a trial that occurred without the error, a guilty verdict would surely have been rendered, but whether the guilty verdict actually rendered in this trial was surely unattributable to the error.’ ” State v. Leteve, 237 Ariz. 516, ¶ 25, 354 P.3d 393, 401 (2015), quoting Sullivan v. Louisiana, 508 U.S. 275, 279, 113 S.Ct. 2078, 124 L.Ed.2d 182 (1993). An appellate court making this harmless-error determination “does not ... ‘become in effect a second jury to determine whether a defendant is guilty.’ ” Neder v. United States, 527 U.S. 1, 19, 119 S.Ct. 1827, 144 L.Ed.2d 35 (1999), quoting Roger J. Traynor, The Riddle of Harmless Error 21 (1970). Nonetheless, the appellate standard parallels the evidentiary standard required to convict: “We must be confident beyond a reasonable doubt that the error had no influence on the jury’s judgment.” State v. Bible, 175 Ariz. 549, 588, 858 P.2d 1152, 1191 (1993). Beyond these broad statements, there is no bright-line rule for what constitutes harmless error. Id. Not surprisingly, to the extent that the harmless error doctrine has been applied outside an analytical frame*508work, it has been criticized as conclusory. See, e.g., D. Alex Winkelman et al., An Empirical Method for Harmless Error, 46 Ariz. St. L.J. 1405, 1412-14 (2014); see also Charles S. Chapel, The Irony of Harmless Error, 51 Okla. L. Rev. 501, 505-06 & n,29 (1998) (criticizing harmless error review and listing other critical journal articles).
¶ 8 In Bible, our supreme court favorably cited Weinstein’s Evidence for the variety of factors frequently considered by courts. 175 Ariz. at 588, 858 P.2d at 1191. They are4:
(1) Whether the other evidence is overwhelming;
(2) Whether the erroneously excluded or admitted evidence would have been primary evidence or material fact;
(3) Whether the party was able to present the substance of the claim or defense;
(4) The cumulative effect of all errors;
(5) Whether erroneously admitted or excluded evidence is merely cumulative of similar evidence already received;
(6) Whether the relevant jury instructions were appropriate and useful;
(7) Whether the jury argument was based on tainted evidence; and
(8) The prejudicial effect of erroneously admitted or excluded evidence.
1 Jack B. Weinstein & Margaret A. Berger, Weinstein’s Federal Evidence § 103.41[5] (2d ed. 2016) (hereinafter “Weinstein”). This list incorporates the approaches of both parties and allows a more objective basis to review for harmless error. Thus, we consider these factors in light of all of the evidence below. See Bible, 175 Ariz. at 588, 858 P.2d at 1191 (harmless error review is a case-specific factual inquiry based on all evidence). Not all factors will apply, nor will they all carry the same weight in a particular case.5
Overwhelming Evidence
¶ 9 We begin with the issue of overwhelming evidence because it alone may be dispositive. For example, in Bible, the improper admission of DNA evidence was found harmless because the other evidence of guilt was “far beyond overwhelming.” 175 Ariz. at 588-89, 858 P.2d at 1191-92. The court in Bible detailed six paragraphs of additional evidence that linked Bible to the abduction and murder of the victim. Id.
¶ 10 The state6 contends there is overwhelming evidence of Romero’s guilt because his cell phone was found at the scene of the shooting; the victim was in the contacts list of the phone; two of the men seen fleeing the scene were described as Hispanic; one of the fleeing men referred to another as “Joe”; the description of the truck at the scene was similar to a Ford Ranger Romero had driven; and, Romero discarded a Glock one month after the shooting when he was being followed by police. To determine the significance of the evidence, we start with the offense charged and then address specific facts.
¶ 11 Romero was indicted for first-degree murder, but the jury was not instructed on accomplice liability; rather, it was instructed only on the lesser offenses of second-degree murder and manslaughter. Moreover, the state only argued to the jury that Romero fired the fatal bullet and explicitly rejected before the court an alternative liability theory. The limited charge and instructions, and *509the absence of argument to the contrary, precluded alternative theories of criminal liability that might have included shots fired by other persons present at the scene. See, e.g., State v. King, 226 Ariz. 253, ¶ 16, 246 P.3d 938, 943 (App.2011) (criminal liability may be based on accomplice theory if state proves defendant aided or facilitated commission of offense by principal). Thus, it was not sufficient to prove only that Romero was present and may have fired a gun at the scene.
¶ 12 The evidence unrelated to the Glock supports the contention that Romero was at the scene, but it is not overwhelming. For instance, the eyewitnesses differed as to whether there were two or three men running from the scene. The eyewitness who was able to estimate the height of the men initially described them as five feet, seven inches to five feet, eight inches, but added several inches to the estimate in his trial testimony. Romero is six feet, one inch to six feet, two inches. The same witness identified the truck as a Ford Ranger or a Mazda pickup. Further, the evidence did not show that Romero was the exclusive driver of the Ford Ranger; rather, it suggested he was merely one of its users. Finally, the name reference could have been mistaken for “Joel,” the victim’s roommate and the initial suspect in the case. Notably, no witness knew Romero or could identify him at trial as a person at the scene.
¶ 13 Even if we assume the eyewitness testimony and the location at the scene of a cell phone used by Romero suggests he was one of the two or three men running after four or five shots were heard, it was the state’s burden to prove Romero was the person who actually fired the fatal shots. Moreover, although two bullets were recovered from the victim’s body, there was no evidence linking those bullets to specific shell casings. Thus, the match of a Glock—separately and independently linked to Romero—to the shell casings found on the ground was the strongest evidence that he fired the shots. Removing from consideration a factual conclusion of a match, the remaining evidence was arguably sufficient to survive a Rule 20 motion, but the state’s burden is greater: it must show overwhelming evidence. State v. Anthony, 218 Ariz. 439, ¶ 41, 189 P.3d 366, 373 (2008) (state’s burden exceeds “whether the jury was justified in its verdict”). Finally, we note that Romero’s first trial resulted in a hung jury. See State v. Rich, 184 Ariz. 179, 181, 907 P.2d 1382, 1384 (1995) (noting hung jury in conclusion error not harmless); see also Cobb v. State, 283 Ga. 388, 658 S.E.2d 750, 753 (Ga. 2008) (noting “prior hung juries are a factor supporting a finding of harmful error”); State v. Edwards, 131 Wash.App. 611, 128 P.3d 631, ¶¶ 16-17 (Wash.Ct.App. 2006) (noting previous hung jury in considering whether untainted evidence overwhelming), We are skeptical the prior jury would have been unable to reach a verdict if the evidence was indeed as “overwhelming” as the state maintains. On this record, we cannot conclude there was overwhelming evidence that Romero fired the gun that killed the victim. See State v. Lehr, 201 Ariz. 509, ¶¶ 31-43, 38 P.3d 1172, 1181-83 (2002) (preclusion of cross-examination and defense expert testimony criticizing DNA testing not harmless where DNA was key evidence).
Primary Evidence
¶ 14 In determining whether evi-dentiary errors are harmless, courts also consider whether the error involved the admission or exclusion of primary evidence. Compare State v. Ontiveros-Loya, 237 Ariz. 472, ¶ 22, 352 P.3d 941, 947-48 (App.2015) (error not harmless where erroneously admitted photographs “strongest evidence produced at trial” that defendant possessed deadly weapon), with State v. Nelson, 146 Ariz. 246, 248-49, 705 P.2d 486, 488-89 (App. 1985) (any error in admitting intoxilyzer test results harmless as to conviction for unlawful flight from pursuing law enforcement vehicle). Here, Haber’s proposed testimony, generally critiquing the science of firearms identification, would not have included evidence that Haber had reached a different conclusion than Powell; it was only intended to weaken Powell’s testimony that the gun and shell casings matched, without directly contradicting Powell’s findings. Haber’s testimony therefore was not primary evidence. Cf. Petree v. Victor Fluid Power, Inc., 887 F.2d 34, 41-42 (3d Cir.1989) (erroneous preclusion of rebuttal evidence that “totally contradicted” defense “star witness” not harmless).
*510Opportunity to Present Claim or Defense
¶ 15 Whether an error is harmless may also be considei’ed in the context of a party’s ability to present the substance of his claim or defense. See, e.g., Gaston v. Hunter, 121 Ariz. 33, 52-53, 588 P.2d 326, 345-46 (App.1978) (exclusion of some expert opinions harmless where substance elicited at other times in trial). The substance of Romero’s claim regarding the match of the gun was that not enough is known about the uniqueness of gun toolmarks7 to warrant reliance on a match by an examiner. Romero impeached the foundation of Powell’s opinions in cross-examination using a report from the National Academy of Sciences (NAS) that criticized “the lack of a precisely defined process” in toolmark analysis, which makes difficult “well-characterized confidence limits.” Nat’l Research Council, Strengthening Forensic Science in the United States: A Path Forward 155 (2009). In fact, Haber’s opinions relied in no small part on the NAS report. Accordingly, we conclude Romero was able to present the substance of his challenge to the reliability of Powell’s opinions.8
Cumulative Effect of Errors
¶ 16 Evidentiary errors may compromise only a small portion of the total evidence or they may be repeated with a significant cumulative effect. See, e.g., Town of Paradise Valley v. Laughlin, 174 Ariz. 484, 487, 490, 851 P.2d 109, 112, 115 (App.1992) (vacating and remanding due to “cumulative effect of the court’s error,” although preclusion of defendant’s testimony alone would not warrant new trial); Coyner Crop Dusters v. Marsh, 91 Ariz. 371, 375, 372 P.2d 708, 711 (1962) (although new trial required because of incorrect jury instructions, other errors also considered because “cumulative effect” could result in unfair trial). In criminal cases, however, Arizona rejects the “cumulative error doctrine” outside the context of prosecu-torial misconduct claims. State v. Hughes, 193 Ariz. 72, ¶ 25, 969 P.2d 1184, 1190-91 (1998). Because there is no claim of prosecu-torial misconduct, this factor does not apply.
Cumulative Evidence
¶ 17 The significance of evidence erroneously admitted or excluded may depend on whether it is more of the same type of evidence properly admitted in the case. See, e.g., State v. Williams, 133 Ariz. 220, 226, 650 P.2d 1202, 1208 (1982) (erroneous admission of hearsay harmless where cumulative to testimony at trial). Cumulative evidence supports a fact “otherwise established by existing evidence”; that is, it is not enough to be simply corroborated by other evidence, and it cannot be the very issue in dispute. State v. Bass, 198 Ariz. 571, ¶ 40, 12 P.3d 796, 806 (2000).
¶ 18 That Romero was able to impeach Powell using the NAS report does not mean Haber’s opinions would have been cumulative. For example, his proposed testimony provided context for the report’s criticism of firearms identification methods, explaining how the accuracy of the field could be challenged. Haber also could have responded directly to statements made by Powell. For example, at the conclusion of Powell’s testimony, a juror submitted a question as to whether a different examiner with similar training and experience would reach the same conclusion regarding the shell casings, and Powell said he was “completely confident.” Haber’s testimony would have addressed the scientific basis for that confidence. Likewise, Haber would have been able to answer jury questions. Ariz. R. Crim. P. 18.6(e). In this case, the expert testimony would have augmented the cross-examination about the report, rather than repeating it. Cf. State v. Ray, 123 Ariz. 171, 173 & n.1, 598 *511P.2d 990, 992 & n.1 (1979) (erroneous admission of preliminary hearing transcript not harmless where testimony contained therein not cumulative).
Jury Instructions
¶ 19 Errors also may be vitiated or exacerbated by jury instructions. See Weinstein § 103.41[5]; cf. State v. Schroeder, 167 Ariz. 47, 50-51, 804 P.2d 776, 780-81 (App. 1990) (expert’s improper testimony regarding credibility of victim cured by court’s admonition and instruction to jury regarding its role in determining credibility). Here, the jury was instructed that it was “not bound by any expert opinion,” and could choose how much weight to give the opinion. Although this allowed the jury to choose not to credit Powell’s testimony, we cannot say it either vitiated or exacerbated the error caused by the preclusion of Haber’s testimony.
Jury Arguments
¶ 20 The effect of erroneous rulings may also be compounded by reference to missing or improper evidence in arguments to the jury. See Mueller v. Hubbard Milling Co., 573 F.2d 1029, 1037 (8th Cir.1978) (error not harmless where counsel emphasized erroneously admitted evidence in closing argument); see also State v. Watkins, 126 Ariz. 293, 299, 614 P.2d 835, 841 (1980) (error harmless where erroneously admitted evidence not mentioned in opening or closing arguments). Here, the prosecutor argued, “[Y]ou cannot take [the criticisms in the National Academy of Science book] as some kind of evidence, because there is no evidence from this courtroom, from that witness stand that actually challenges firearms analysis.” Although the argument was not improper in view of the preclusion ruling, it highlights the fact that the absence of direct evidence challenging Powell’s opinion was sufficiently important to the state’s position that it argued to the jury the NAS report should not be used to impeach Powell. We conclude the state’s emphasis on the defendant’s lack of “witness stand” evidence supporting the NAS report exacerbated the error of Haber’s preclusion.
Prejudice
¶ 21 Prejudicial effect is implicit in all factors, but also stands as a catch-all category when consideration of a particular factor would not otherwise apply. Here, the preclusion of a defense witness who would have challenged the testimony of the state’s expert carried additional prejudice, because “ ‘science’ is often accepted in our society as synonymous with truth.” Bible, 175 Ariz. at 578, 858 P.2d at 1181, quoting 1 Morris K. Udall et al., Arizona Practice: Law of Evidence § 102, at 212 (3d ed. 1991).
Summary and Disposition
¶ 22 Looking at “the trial record as a whole,” United States v. Hasting, 461 U.S. 499, 509, 103 S.Ct. 1974, 76 L.Ed.2d 96 (1983), the evidence of guilt was not overwhelming, Haber’s testimony was not cumulative, no jury instruction ameliorated the error, and the arguments to the jury compounded the error. We cannot say beyond a reasonable doubt that the preclusion of Haber’s testimony had no influence on the jury’s verdict. Therefore, the error here was not harmless and we must reverse. Romero’s conviction and sentence are reversed, and this case is remanded to the superior court for further proceedings.
. Daubert v. Merrell Dow Pharm., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993).
. Haber testified at the Daubert hearing.
. The Arizona Rules of Evidence generally correspond to the Federal Rules of Evidence, but nonetheless stand on an independent basis and sometimes "deliberately depart” from the federal rules. Ariz. R. Evid., prefatory cmt. to 2012 amendments. Similarly, because our rules of evidence are founded in state procedural law, "federal court decisions interpreting the federal rule are persuasive but not binding with respect to interpreting the Arizona rule.” Id.
. We have changed the order of the factors to facilitate our analysis.
. Our dissenting colleague objects to the categorization of factors as creating a fragmented approach which he suggests departs from the standards our supreme court has established. We respectfully disagree. Bible first cited Weinstein for the factors potentially relevant in a harmless error analysis. 175 Ariz. at 588, 858 P.2d at 1191, Moreover, each elucidated factor is based on or consistent with Arizona case law. In contrast, the dissent’s concluding reliance on State v. Anthony, 218 Ariz. 439, ¶ 39, 189 P.3d 366, 373 (2008), follows a summary description of one factor we address in detail—but which is separate from additional factors not limited to evidence of guilt. We also observe that although the court in Anthony provided a very detailed recitation of the circumstantial evidence supporting the murder conviction, its conclusion that the evidentiary error was not harmless involved weighing the impact of the improper evidence. Id. ¶¶ 3-23, 40-42.
.The dissent enlarges the discussion of several facts that are referred to briefly by the state. To maintain context, we discuss them here and with singular attribution to the state's position.
. A "toolmark” is the impression left on a softer surface when it comes into contact with a hard object. In this context, it refers to imprints left by the gun’s firing pin and related parts on the bullet casings. See generally United States v. Monteiro, 407 F.Supp.2d 351, 359-61 (D.Mass.2006).
. Romero argues for the first time in his supplemental brief on remand that Powell changed the basis of his opinion between the two trials, by testifying at the first trial that he relied on the “fingerprint" of the breech face of the gun on the cartridge and, at the second trial, that he was referring to "shearing” caused by up-and-down motion of the cartridge against the firing pin aperture at the second trial. A review of the context of his testimony, however, indicates he was referring to the same marks at both trials.